The Trade Desk (TTD) - A Champion of the Open Web

Bristlemoon's exploration of the murky, complex world of ad tech

Welcome to the fifth Bristlemoon Capital stock write-up. It has been five months since publishing our first post, and this report on TTD will mark more than 80,000 words of content being published cumulatively over that period. We recently made the decision to monetize this publication, and we are immensely grateful for the many subscribers who have already converted to paid subscriptions.

Please see the following links for our previous reports on EYE, RH, Copart, Match Group, as well as our appearance on the Business Breakdowns podcast discussing Match.

Bristlemoon readers can also access some of the resources we used to research TTD via the links below:

• Tegus – Bristlemoon readers can enjoy a free trial of the Tegus expert call library via this link

• Koyfin – Bristlemoon readers can get 20% off a Koyfin subscription via this link

Table of Contents

• Introduction & company history

• Programmatic advertising and the ad tech supply chain

  • Programmatic buying models

  • Ad tech supply chain

  • Walled gardens vs the open web

• Business overview

  • A champion of the open web

  • Financial performance

  • Platform & technology

• Competitive differentiation

  • Scale

  • Independence

  • Close relationship with ad agencies

  • Transparency & control

For paid subscribers:

• Growth drivers and other opportunities

• Connected TV

  • Overview

  • Market fragmentation

  • The shift to programmatic

  • Programmatic fees

  • Programmatic CPMs

  • Will CTV be 100% biddable?

  • TTD’s market share

• Retail Media

  • Overview

  • Onsite vs offsite

  • In-housing of retail media

  • TTD’s offsite retail media partnerships

  • How big is the opportunity?

• Take rate sustainability

  • Overview

  • Impact of CTV mix

  • 3^rd party cookie deprecation

  • OpenPath and Supply Path Optimization

• Unified ID 2.0 and the deprecation of cookies

  • UID2

  • Privacy issues

  • Signal loss

  • Impact of cookie deprecation

• Valuation considerations

Key Takeaways

• The Trade Desk is the largest independent Demand Side Platform in the programmatic advertising industry, used by advertisers to automate the process of buying ads on the open web. Led by enigmatic and promotional co-founder Jeff Green, the company has delivered phenomenal growth and a mind-blowing 40x share price return in just over seven years as a public company.

• The programmatic ad tech supply chain is highly opaque and rife with conflicts of interest and fee gouging. The Trade Desk built its reputation on independence, transparency, and exclusively serving the interests of its ad buying clients. A prescient decision to align with the advertising agency holding companies early on allowed The Trade Desk to turbocharge its growth in an industry where scale begets scale. As the largest independent DSP, the company is poised to be the key beneficiary of industry consolidation.

• The Trade Desk is overweight two of the fastest growing channels in digital advertising: connected TV and retail media. CTV is rapidly taking share from linear TV in both share of time spent and share of ad spend, and management spin an appealing narrative around the company’s opportunity in CTV advertising. We challenge management’s claims of perfect market fragmentation, an inevitable shift to fully programmatic ad sales, and higher industry CPMs. Instead, we find a market that will be dominated by a handful of powerful players capable of erecting walled gardens, a structural supply-demand imbalance favoring ad sellers over ad buyers, and industry economics that will be less favorable going forward.

• Retail media is a long-term opportunity and one that we believe is even more challenging for The Trade Desk to exploit given its lack of required supply-side capabilities, a lack of 1^st party data, retailer pushback to ad tech fees, and a changing privacy/cookies landscape that risks degrading the efficacy of offsite retail media advertising.

• We believe the company’s unerringly stable take rate of ~20% since IPO may come under increasing pressure in the coming years with headwinds from an evolving CTV mix and Google’s planned deprecation of 3^rd party cookies in the Chrome browser, which will impact (very) high fee data sales. Management is attempting to mitigate fee pressure through supply path optimization, a.k.a. bypassing other intermediaries in the supply chain to retain more fees for itself, which every other prominent ad tech intermediary is also doing. The company is also trying to offset the loss of signal from cookie deprecation by implementing its proprietary Unified ID 2.0 identifier solution, though we are not confident such alternative deterministic IDs are scalable or will pass regulatory scrutiny in the long run. 

• Ultimately, we find The Trade Desk to be a fantastic product, a sound business with good prospects if ignoring the hype, but an unattractive stock that is priced to perfection with no margin of safety for the myriad of risks that shareholders face.

Introduction

If a movie were to be made about digital advertising, Google would be the Roman Republic and Jeff Green would be Spartacus. Green, the enigmatic co-founder of The Trade Desk, would lead his band of rebels in a fight for the freedom of the Open Internet against the tyranny and oppression of the Google ad empire. Through a combination of prescient early decisions, a healthy dose of right-time-right-place, and an unwavering commitment to serve the interests of its core clients in an industry that is rife with conflicts of interest, The Trade Desk has achieved something truly remarkable: a 40-bagger in seven years as a public company.

This deep dive marks our first foray into the murky world of advertising technology (ad tech). We found the industry to be opaque, unnecessarily complicated, and challenging for non-practitioners to fully understand all the nuances. As the poster child of independent ad tech and a current market darling, The Trade Desk seemed like the logical place to start. We caution readers again that this report presents our initial research into ad tech and The Trade Desk; we are not authorities on either and shouldn't be considered as such. We invite readers to partake on a journey with us as we learn about both industry and company. As always, any pushbacks or qualifications to our views and conclusions are greatly appreciated.

The story so far

Jeff Green’s entrepreneurial beginnings

The story of The Trade Desk is inextricably linked to the story of the company’s co-founder, Jeff Green. Green, a Utah native, earned a Bachelor of Arts degree from Brigham Young University in 2001 and then later a degree in marketing communications from the University of Southern California[1]. It was during his studies that Green gained experience as a digital media buyer at an agency during the 2000 dot-com crash[2]. Green later recounted how these formative experiences fueled his passion for digital advertising:

“I fell in love with digital advertising then and there. I loved the fusion of left brain and right brain. I got be to be creative and imagine a new hypothesis to make advertising better and then we’d get to look at the data and validate or disprove the idea. Back then I realized the advantages of data-driven digital advertising, which allowed advertisers and marketers to target and also measure ads like never before. I’ve basically spent most of the last 20 years trying to make that more pervasive.”[3]

Following his tertiary studies, Green started his professional career as a technical account manager within the MSN division of Microsoft, based in Salt Lake City. It wasn’t too long after this that Green started his own business. In 2003, Green founded a company called AdECN, one of the world’s first online ad exchanges. AdECN was akin to a stock market for digital ads (note that Jeff Green got his securities broker license in his twenties and had previously worked in the investment division of an insurance company).[4]

In 2007, four years into his journey running AdECN, Jeff Green’s career had come full circle, with Microsoft agreeing to acquire his company for an undisclosed sum. AdExchanger speculated on the purchase price, writing about the company being acquired “for what some suggest is in the $50-75 million range”[5]. Green continued to run AdECN within Microsoft, and he took on additional responsibilities including overseeing the reseller and channel partner business which included monetization of Facebook ads, Fox Sports, MSNBC, Hotmail and several other internet sites[6].

The Trade Desk – the Goldman Sachs of online advertising

It was during this second stint at Microsoft that Jeff Green met David Pickles, his soon-to-be co-founder and CTO. In 2009, Jeff Green and David Pickles co-founded The Trade Desk. Within the confines of Microsoft, Green was hamstrung in his ability to enact his ideas for the future of digital advertising technology. By starting his own company (again), Green reclaimed agency to continue building without “the bureaucracy and internal politics”[7] that came with being nestled in a bigger firm. 

In a February 2023 Invest Like the Best podcast interview, Jeff Green explained his motivations for starting The Trade Desk:

“In my previous company, I was very frustrated by the fact that online advertising was transacted the same way traditional advertising was, which was over martini lunches. If you watch Mad Men, that's how it was at the beginning of the Internet, too, more casual clothes, but same process. It just frustrated me why couldn't the advertising ecosystem look more like the equities markets. Why couldn't we have a centralized exchange? In my 20s, I was young enough, naive enough, dumb enough, ambitious enough to try to create that and change all of it. And I happen to be lucky enough to be one of the people that perpetuated that movement and created that. So there were a number of us, I wasn't the only one. But there were a number of us that created ad exchanges that now have changed advertising. So those are at the center of the ecosystem.”

In Green’s previous role at AdECN, he noted a lack of sophistication in the ad buyers. Despite the creation of ad exchanges, ad buyers still defaulted to paying $0.50 for every impression. They would bid the same price for every impression in a world where every impression was clearly not created equal. Green’s frustration with the underutilization of ad tech’s capabilities prompted him to start The Trade Desk and exclusively represent ad buyers to maximize the value of their ad dollars within the digital ecosystem.

Green once told socalTech that “[w]hat we’re building now is a system much more like the Goldman Sachs of online advertising”. The platform would offer a marketplace enabling buyers to place specific ads in-front of a precisely targeted audience[8], utilizing data to automate these decisions in real-time.

Taking The Trade Desk public

The Trade Desk raised $2.5 million from Founder Collective and IA Ventures in a seed round in March 2010[9]. In addition to the two lead investors, there were a number of strategic investors who contributed capital in a second and final tranche for the seed round. These investors included Jerry Neumann, who previously ran the interactive media arm of Omnicom, and Josh Stylman, who was co-CEO and Managing Partner of Reprise Media, a division of Interpublic[10].

In a 2021 article covering the launch of the company’s internal venture capital arm, TD7, it was disclosed that the “TD7” name derived from the $7 million in venture capital it took for The Trade Desk to become profitable[11], implying an additional $4.5 million of venture capital from the strategic investors. Co-founders Green and Pickles asked for an amount of money needed to get the business to profitability, and The Trade Desk was able to achieve profitability in 2010, the company’s second year of operation.

The Trade Desk completed its IPO on September 21, 2016, trading on the Nasdaq under the ticker TTD. The company went public at $18.00 per share and finished its first trading day at a price of $30.10 per share, a 67% increase. From TTD’s S-1 filed in 2016, the company produced $113.8 million of revenue in 2015, a 156% increase over the $44.5 million of revenue produced in 2014. Growth in TTD’s adjusted EBITDA was even more explosive. The company generated $39.2 million of adjusted EBITDA in 2015, growth of 589% over the $5.7 million of adjusted EBITDA produced in 2014. The company was notching up astonishing growth rates and investors were scrambling to own this business.

The company’s share price performance since the IPO puts TTD in the compounder hall-of-fame. If we adjust the IPO price for the ten-for-one split of the company’s common stock that occurred on June 17, 2021, we can calculate a $1.80 split-adjusted IPO share price. Given the share price is currently around $74, this equates to a 41x increase in value since the September 2016 IPO. This works out as a 64% CAGR over seven years. If we were to adjust for the day-one share price pop, and instead use the $3.01 split-adjusted price in our calculation, this still produces a 53% CAGR.

Source: Koyfin (Bristlemoon readers can get 20% off a Koyfin subscription via this link)

Before we can get into what makes The Trade Desk special and what the future potentially holds, we must first establish a foundational understanding of the programmatic ad tech supply chain in which the company operates. Readers familiar with the ad tech ecosystem can feel free to skip the following section.

Programmatic advertising and the ad tech supply chain

Programmatic advertising refers to the use of technology and algorithms to automate the process of buying digital ad impressions. It offers improvements over more traditional manual buying methods in terms of cost savings, time savings and better targetability of ads. The term “programmatic” refers to not one but a range of buying models with varying levels of automation, targeting, data activation and control retained by the advertiser and publisher. The price at which ads are bought and sold at is typically referred to as CPM (cost per mille or 1,000 impressions) or some other cost-per-unit that can typically be converted back to an effective CPM.

Programmatic buying models

Source: Bristlemoon Capital

Programmatic Guaranteed

The most basic yet highest priority form of programmatic buying is Programmatic Guaranteed (PG), which are essentially digitalized versions of direct deals. PG deals are typically negotiated as one-on-one deals between an advertiser and publisher, have a fixed price and a guaranteed volume of impressions. The only automated decision is whether or not to buy each ad impression as they arise, noting that the advertiser must buy (and the publisher must serve) the agreed-upon volume of impressions for the campaign; everything else is negotiated beforehand.

The benefit to the advertiser is that it receives a guaranteed volume of premium inventory (i.e., the most engaged audiences) for a campaign at a fixed price but surrenders a lot of flexibility in terms of using data to target specific audiences. The benefit to the publisher is that it retains the greatest control over the ads served on its properties and achieves a favorable fixed price, but risks underpricing its inventory relative to an open auction with price discovery. Because the advertiser and publisher deal directly, the risk of ad fraud is greatly reduced. PG is typically only used by “premium” publishers with the most sought-after inventory that can fetch a high negotiated price.

Preferred deals

Next in the order of priority are preferred deals, which are identical to PG except the impressions are not guaranteed. Instead, the buyer gets a “first look” at premium ad impressions and can choose to buy or pass at the negotiated price. Because there is no obligation to buy, there is more scope for the advertiser to activate audience data to inform the buying decision even though the pricing remains fixed. Impressions that are not bought are passed down the priority ranking to private or open auctions.

Private Marketplace

The first of the auction models is the Private Marketplace (PMP), which as the name implies is available by invitation only. A publisher of premium inventory invites a group of advertisers into a PMP, and the advertisers use real-time bidding tools to participate in a private auction. The invite-only nature of a PMP makes it safer for both publishers and advertisers. Publishers invite only trusted brands and demand partners into the auction and thus retain some element of control over what ads are shown where, and advertisers know their ads will be shown in brand-safe environments with minimal ad fraud. CPMs may also be higher as: 1) there is a real-time auction albeit with a select group of bidders; and 2) the publisher, advertisers and their ad tech partners can share and leverage 1^st and 3^rd party data in a safe environment to allow better audience targeting, which can drive higher CPMs and higher return on ad spend (ROAS).

For the above three buying models – PG, preferred deals and PMP – the buyers and sellers are known to each other and can negotiate the terms of their exchange and build advertiser-publisher relationships over time. This has implications for ad tech partners as we will discuss later in the report.

Open auction

Finally, there is the open auction or open exchange, which is where real-time bidding occurs in an open marketplace of publishers and advertisers. Publishers will specify certain parameters around their inventory (e.g., format, ad type), set floor prices and send their bid requests out to the open market for any eligible advertiser to bid on. Advertisers can set up automated campaigns that will automatically bid on any bid requests that match the advertiser’s criteria.

The publisher does not know who the winning bidders are and what kind of ads will appear on their properties. They also face an elevated risk of data leakage as each bid request sent carries a snippet of valuable 1^st party publisher data that can be scraped by malicious actors. Advertisers do not always know what kind of inventory their automated campaigns are bidding on or what content their ads will appear next to. They are also subject to a heightened risk of ad fraud such as bidding on zombie audiences or Made-for-Advertising (MFA) inventory with zero value.

Open auction CPMs can be mixed as there is an unknown quantum and quality of bidders, though open exchanges generally have a much larger supply of potentially lower quality impressions compared to the other programmatic channels discussed above. It is for these reasons that premium publishers with premium inventory typically reserve open auctions for any remnant inventory not sold through other means.

The ad tech supply chain

In order to facilitate the automated buying and selling of ads, an ad tech ecosystem has developed over the past three decades to serve digital advertisers and publishers. To say that the ad tech supply chain is needlessly complicated is a gross understatement. An ad dollar leaving an advertiser’s bank account will typically pass through 4-5 intermediaries before reaching the publisher’s pocket depending on the programmatic buying model, not including the numerous vendors to those intermediaries who also take their cut of fees. By the time the original ad dollar reaches the publisher, it has typically shrunk to less than seventy cents.

Below we provide a brief introduction to the key participants of the ad tech supply chain.

Source: Redburn. Note: DSP = demand-side platform; SSP = supply-side platform

Advertiser – the party with the ad budget looking to buy ad impressions to promote their product, service, or brand.

Agency – an organization that manages the ad buying process on behalf of advertisers. Agencies typically provide the hands-on-keyboards workforce to run ad campaigns for their clients. They can exert significant influence on how their clients allocate or spend their ad budgets. “Ad buyer” can refer to the agency or the advertiser if buying direct. Agencies are either independent or subsidiaries of the large agency holding companies (“holdcos”) – WPP, Omnicom, Publicis, IPG, Dentsu and Havas.

Publisher – owner or operator of media properties with advertising inventory (i.e., ad space) that can be sold to ad buyers. Can range from “traditional” media operators such as newspapers, cable networks and connected TV platforms through to special interest websites, e-commerce marketplaces and Facebook.

Demand Side Platform (DSP) – ad tech vendor to advertisers and agencies, providing the automated ad buying tools to bid on auctions, run and optimize ad campaigns, and connect into supply partners be they publisher direct, an ad exchange or an SSP. TTD is a pureplay DSP.

Supply Side Platform (SSP) – ad tech vendor to publishers, providing the technology for publishers to manage and automatically sell their ad inventory. An SSP manages the bidstream – the stream of bid requests that emerge as viewers navigate around a publisher’s property – and finds optimal placements for ads to maximize the publisher’s yield. An SSP integrates with demand partners such as ad exchanges or DSPs to bring advertisers to the publisher’s audience.

Ad exchange – a marketplace where ad buyers, DSPs, SSPs and ad sellers connect to transact through automated auctions. Auctions are typically conducted on an ad exchange but can also be run directly by an SSP if it has the capabilities – for example in a private marketplace auction. Many SSPs also own an ad exchange and ad server, with Google Ad Manager (SSP), Google AdX, and Google Ad Server being the most prominent example.  

Ad server – ad servers sit at both ends of the supply chain. Advertiser ad servers store ad creatives (the images, audio and video that comprise an ad) and provide some measurement capabilities. Publisher ad servers have a more prominent role in ad tech as they determine the winner of an auction, which ads are placed where, and handle some ad targeting and measurement functions.

Other vendors in the ad tech ecosystem outside the remit of this TTD deep dive include (but not limited to):

• Data management platforms which collect or scrape 3^rd party data to be sold to advertisers via DSP integrations;

• Measurement and verification partners that measure campaign performance metrics such as reach, brand lift, conversions etc, and independently verify ads for viewability, brand safety, ad fraud etc.; and

• Data clean rooms and customer data platforms which manage sensitive 1^st party data and personally identifiable information in a safe environment, allowing this valuable data to be used to build and target audiences.

Below is a sample of leading ad tech ecosystem participants:

Source: Morgan Stanley

Walled gardens and the open web

The digital advertising landscape can be separated into two continents – the walled gardens and the open web. The walled gardens, as the name suggests, are platforms or media properties that are largely closed off to external participants in the ad tech supply chain. The media owner typically controls the demand side tools that advertisers require to bid for ad impressions on its owned & operated (O&O) properties, the supply side tools for managing inventory, bid requests and auctions on its properties, and the ad server that serves the winning ads to the properties’ audience. The most prominent walled gardens (ex-China) are Google and Meta, and tier 2 walled gardens include Amazon, TikTok, Snap and X (Twitter) among others.

The key criteria for walled gardens are 1) having massive, internet-scale audiences that can attract ad buyers directly, and 2) having highly valuable 1^st party data that is only available within the walled garden. The open web, on the other hand, represents the millions of other publishers on the Internet vying for a share of digital ad dollars. These publishers range from household names such as Disney, Walmart and The New York Times, through to the long tail of unheard-of websites with minimal traffic.

Jounce Media[12] forecasts open web programmatic advertising to reach ~$88 billion in 2023, a CAGR of 5% since 2017, compared to over 30% CAGR for the walled gardens. Google is so dominant in the ad tech industry that it serves ~40% of all ad dollars transacted on the open web, with the remainder mainly being served by the independent ad tech supply chain. TTD is the largest independent ad tech provider with $7.7 billion in gross advertising dollars flowing through its buying platform in 2022. TTD has just under 10% share of the market.

Source: ANA, Jounce Media. Walled Gardens includes Meta (Facebook & Instagram), YouTube, Amazon properties, and “Challenger” gardens TikTok, Twitter, Snap, LinkedIn, Pinterest etc.

Source: Jounce Media; Bristlemoon Capital

A champion of the open web

The Trade Desk provides a self-service, cloud-based programmatic ad-buying platform. TTD’s clients are ad buyers. These ad buyers use TTD’s platform to plan, manage, optimize and measure their digital advertising campaigns. The platform allows clients to execute ad campaigns across various advertising channels including video and connected TV, display, audio, digital out of home, and social. It also allows ad campaigns to span across a range of devices including computers, smartphones, smart TVs and streaming devices. As a pure-play DSP, TTD doesn’t own or operate any of its own ad inventory. Instead, it obtains a supply of omnichannel inventory from over 100 directly integrated ad exchanges and SSPs.

TTD has positioned itself as the go-to independent open web ad buying platform for large, sophisticated advertisers typically represented by ad agencies. As of December 2022, the company has over 1,000 clients, consisting primarily of advertising agencies under the large agency holdcos and tier 2 independent agencies, with each agency likely representing a multitude of advertisers.

Clients are defined as parties that have signed agreements with the company and spend more than $20,000 on the platform, though we have anecdotally seen minimum spend requirements of over $100,000 per month for prospective direct clients. Smaller advertisers and agencies are required to go through resellers. Management have stated that they are not interested in onboarding sub-scale clients and are more focused on growing share of ad spend at existing clients. TTD’s client retention rate has exceeded 95% for 2022, 2021 and 2020.

Source: Redburn; Company filings

TTD’s clients are diversified across verticals, with no vertical representing more than a high-teens percentage of total gross spend on the platform. This derisks the revenues to an extent as TTD is not overexposed to any single industry cycle, though as we saw in Q3 2023, the company is not immune to broader weakness in the digital ad market, particularly within the CTV channel.

Source: Bristlemoon Capital; Company filings

Management doesn’t disclose quantitatively the gross spend by channel mix, but we know video is the largest channel and CTV is the vast majority of video. CTV is the fastest growing channel and largest opportunity for TTD, followed by retail media which is more nascent for TTD but potentially just as large an opportunity if one believes in management. We will discuss both CTV and retail media in detail later in the report given their importance to the TTD thesis. The mobile web and display channels are the more legacy channels for TTD and should continue to decline in importance over time. Redburn estimates the display channel may have even declined slightly into 2023 based on qualitative comments around channel mix, with growth largely driven by CTV and to a lesser extent retail media. Finally, audio and digital out of home are fast-growing emergent channels but relatively small for TTD and unlikely to move the needle for shareholders.

Source: Redburn estimates; Bristlemoon Capital estimates (based on qualitative comments)

Financial performance

TTD’s unique technology and strong competitive differentiation has enabled it to deliver outstanding financial performance. Since the company’s IPO in 2016, it has compounded gross spend and revenue at a 38% CAGR, handily outgrowing the broader open programmatic market and taking 700 bps of market share between 2017 and 2022. TTD collects a percentage of the client’s media budget as a fee for using its platform. Said another way, TTD has a take rate revenue model. The percentage take rate has remained stable at ~20% of gross spend since IPO. We only mention this briefly here so readers have an idea of TTD’s revenue model, as we have a lot more to say about the take rate later in the report.

Source: Bristlemoon Capital; Company filings

The company has also delivered 800 bps of adjusted EBITDA margin expansion, though the picture is one of deleverage once we factor in share-based compensation.

Source: Bristlemoon Capital; Company filings

We would note here that this report is less quantitative than our previous reports simply because TTD is a very high growth business. We believe qualitatively analyzing the key growth drivers and risks for an industry that is evolving rapidly is more informative for the reader than agonizing over NPVs or earnings power that are highly sensitive to growth rates.

Platform & technology

From our research and conversations with industry experts, we heard mixed opinions regarding the TTD platform capabilities and user experience, with some saying the platform had a steep learning curve and required a lot of training from TTD instructors, and others saying the platform was straightforward to use and the capabilities were somewhat underwhelming for experienced, sophisticated traders. On balance, the views leaned more positive in terms of both capabilities and UI/UX.

One thing that does separate TTD from other DSPs and is perhaps not well highlighted by management or understood by investors, is the platform’s unique bid factoring architecture. Bid factoring is TTD’s original technical claim to fame and allows traders to apply a factor or weight to elements of the bidstream such as ad size, site, device etc. that are more likely to produce the best ad performance. The platform takes these factors and computes the value of impressions in real time and adjusts the CPM bid accordingly.

To understand how this differs to competitors, we need to understand how programmatic ad campaigns are actually constructed. The basic “instruction” for an automated ad buying tool is the line item, which consists of multiple elements such as a budget, a CPM bid, an ad creative, and the targeting criteria. When a request in the bidstream meets the criteria of a line item, a bid is submitted at the preset CPM. Campaigns are essentially a collection of line items. In order to adjust or optimize an ad campaign in flight, line items need to be edited or new items inserted, which can result in campaigns growing to hundreds, thousands or even hundreds of thousands of line items. For an agency trader stretched thin across dozens of campaigns for different clients, managing these thousands of line items can be a nightmare.

TTD’s bid factoring architecture simplifies this by applying factors to elements of a campaign across multiple line items – e.g., adjusting all line items targeting the California region with iOS as the device type up by X% regardless of other targeting criteria. According to one recent Tegus interview with a company former, other DSPs cannot replicate the flexibility of TTD’s bid factoring because they (the competitors, including Google DV360) are built on a legacy line item-based architecture.

“The old DSPs, what they're doing now, they do something called ad grouping, which is similar, but not the same in that, they can let you group [line items] together and it's still very manual for the trader…it's so rigid that you could basically say, because of this ad group or this grouping have similar [performance] traits, I'm only going to bid this much on this group…That's what the old DSPs are copying, but it's different than what The Trade Desk is doing because The Trade Desk can not only just change prices, they can change the geo, they can change the demo target, they can add data into one part of it, they can do different things with those groupings that these other companies can't do.”

But we believe bid factoring may be less of a differentiator now that other DSPs have expanded their custom bidding capabilities. For example, Google DV360’s custom bidding algorithms appear to be more powerful and robust than bid factoring, although they require greater user sophistication to properly utilize the additional functionality. DV360 also offers bid multipliers which is similar to bid factoring as far as we can tell but can’t be used in conjunction with automated bidding algorithms. With TTD’s bid factoring, traders can layer manual bid factors on top of the automated bid optimization tool. We reiterate that not being programmatic ad buyers or ad tech operators, we find it objectively difficult to assess the technical capabilities of one platform against another.

Just to highlight the differing opinions on TTD’s tech once again, another interesting perspective we heard was that historically, TTD was not really a technology company in terms of building ad tech infrastructure and lacks the technical engineering capabilities to do so. Within the ad tech stack, there is ad tech infrastructure that sits at the base, and ad tech applications or platforms that sit on top (such as TTD’s ad-buying platform).

A crude analogy to the financial markets would be the “plumbing” that enables trade execution, clearing and settlement to occur, and the trading “platforms” like Robinhood that sit on top. We are not sure what to make of this view as it could not be corroborated, but it may become pertinent to a “build vs buy” question as TTD seeks to expand further into retail media and internationally where the ad tech ecosystem may not be as developed.

In any case, we do not believe technology moats are a source of durable competitive advantage in tech-forward internet-scale industries such as digital advertising. The topic of moats is a discussion for another day, though this article by Jerry Chen of Greylock Partners is a good explainer as to why technology in and of itself has become a shallower moat. Therefore, we advise investors to not overweight the technical claims made by management teams, especially if they cannot verify first-hand any exuberant assertions that are made. Given TTD’s track record, it is clear that the technology is good enough, and management has been active in upgrading the capabilities and UX of the platform.

The Trade Desk platform components

Here we highlight the key components of The Trade Desk platform.

Solimar is TTD’s revamped user interface that consolidated various tools in the TTD arsenal onto a single platform. The platform offers advanced goals-based media buying and optimization using TTD’s machine learning tool, Koa, advertiser 1^st party data onboarding and a measurement marketplace. We have heard mixed feedback on the ease of use of TTD’s tools, with some saying they are easy enough for a fresh graduate agency trader to use and others saying the tools are catered to sophisticated buyers and can require over 12 months of training to master, but Solimar purportedly simplified the user experience and the visualization of campaign performance.

Kokai (Japanese for “open sea” or “open to the public”), announced in June 2023, is the next iteration of the TTD platform that “incorporates major advances in distributed AI, measurement, partner integrations and a revolutionary, intuitive user experience”. While this is admittedly a cryptic string of buzzwords, it appears Kokai will leverage more machine learning (ML) algorithms across the media buying workflow, especially on the measurement side where it may have been lacking previously. Note however that the new measurement features rely on unproven identity solutions in a post-cookie environment. Part of this platform update includes a new user interface inspired by the Periodic Table which has seemingly received mixed reactions. See this Reddit thread for some entertaining takes on the Kokai launch video, including this zinger from a purported TTD employee:

“Made a throwaway [account] for this. The table is JG's idea. We tried to stop it, but he wants it and it doesn't matter what we say. For the record, most of us think it's bad. It's terrible UX. Of course it's terrible; it was designed by someone who doesn't have an ounce of UX design experience. That table that was shown was literally something JG drew up over a weekend in MS paint it something. Like, it's atrocious. But JG won't meet with anyone but his yes-men who just blow rainbows up his ass, so he probably doesn't know that it's bad and that users will hate it.”

The pervasiveness of Jeff Green’s presence and influence across the company at all levels of decision making was reiterated by several company formers we heard from.

Koa is TTD’s AI tool launched in 2018 that initially provided ML campaign optimization recommendations. It has since added the ability to automatically optimize performance and spend in real time against certain KPIs and create custom audience profiles and segments to target. Koa is also a bid shading tool that aggregates and evaluates past bid data to predict the optimal bid rate for first-price auctions to help clients from overspending.

Galileo is another recently launched tool for advertisers to onboard their 1^st party customer data in a privacy-safe environment for use in campaigns to target new customers. Galileo incorporates integrations with major CRM, customer data platform and other data and clean room providers. The tool uses Unified ID 2.0 (which we discuss later in the report) to match audiences across publishers, devices and channels in a purportedly privacy-safe manner.

Given the importance of data (advertiser 1^st party and 3^rd party data) to programmatic advertising, TTD integrates with over 200 3^rd party data providers which largely helps mitigate TTD’s own lack of 1^st party data due to not being a walled garden or owning/operating any media.

The Kokai user interface – UI design faux pas? Source: The Trade Desk

Competitive differentiation

Now that we are familiar with the ad tech supply chain and TTD as a business, how does the company differentiate itself from competing DSPs?

1) Scale

As the largest independent DSP in the ad tech ecosystem by some margin, TTD benefits from scale in ways that competing independent ad tech vendors simply cannot match. Beyond the obvious benefit of greater fixed cost leverage with more ad spend and revenue on the platform than competitors, TTD is well positioned to be the primary beneficiary and an accelerant of industry consolidation such that scale begets scale.

When it comes to programmatic advertising on the open web, ad demand is finite while the supply of ad inventory is effectively unlimited. This means benefits of aggregating demand are much higher than aggregating supply – that is to say, scale is a defining characteristic of DSPs. Advertisers and agencies want to consolidate their budgets under as few DSP partners as feasible for a number of reasons:

• Higher spend through a single platform not only drives volume discounts but also provides more unified campaign data for evaluating the efficacy of omnichannel ad campaigns;

• Allows for more effective frequency capping (the number of times a user is shown an ad) across publishers and channels;

• Enables wider audience reach with ideally more unique audiences and less overlap;

• Mitigates the risk of bidding against oneself if multiple DSP partners receive bid requests for the same ad impression; and

• Increases the value of 1^st and 3^rd party data by activating available data against a higher volume of ad impressions.

For a large advertiser, it is a no-brainer to allocate a majority of the ad budget to a few of the largest DSPs (e.g. Google DV360, TTD and Amazon DSP) and partner with smaller specialist DSPs for niche campaigns. Publishers, on the other hand, want to be connected to as many sources of demand as possible to not only fill all their ad inventory, but also to achieve the highest CPM for each impression. This difference in incentives largely explains why we have not seen the same extent of industry consolidation on the supply side as we have the demand side.

A recent example of TTD being the primary independent beneficiary of industry consolidation is the June 30, 2023 bankruptcy of MediaMath, a prominent DSP that handled over $590m of ad spend in 2020 and was valued north of $1 billion at one stage[13]. Redburn reports that MediaMath’s largest agency customer, Havas, which likely represented 30% or more of MediaMath’s total spend, transferred 70% of its North American ad spend or just under 30% of its global ad spend on MediaMath to TTD, with the rest substantially all going to Google’s DV360. Assuming the non-Havas budgets shared a similar allocation, this sub-30% capture rate of global spend and 70% capture in North America is significantly higher than TTD’s existing open programmatic share of ~9% and ~16% respectively. This rate of incremental share gain is a strong indicator of TTD’s status as the preferred independent DSP partner for the open web, especially for ad agencies.

One MediaMath former we heard from provided the following anecdote: when he asked a Fortune 50 marketing executive on whether the company would run an RFP for a new DSP following MediaMath’s Chapter 11 filing, the response was “What’s the point? There’s no one out there right now who has global scale like MediaMath did, outside of Google and TTD to an extent.”

In addition to being an industry consolidator, another benefit to scale is the sheer volume of bid requests that TTD’s platform handles. The platform views 13 million queries per second (QPS), which is the number of ad impressions that clients could potentially bid on. This volume of queries is obtained through direct integrations with supply partners and by crawling open exchanges. By viewing more queries than independent competitors, ad buyers using TTD have more opportunities to bid on ad impressions and execute scaled campaigns. And not only does TTD view a huge number of QPS, it is also one of the fastest and most efficient ad buying tools, which matters when auctions take place over milliseconds. Per one Tegus interview with a company former:

“The faster the DSP is at bidding, the more chances it has to win in an auction-based environment…[TTD is] taking billions and billions of queries. So the faster the tool, the more optimization capabilities and all those things together is really what sets them apart…TTD is excellent at optimizing based on efficiency and speed.”

And on why smaller DSPs can’t compete with TTD along this dimension:

“What QPS means is bidding power, bidding power equals engineering, and it also equals data colocation. So the more data centers you have, the more colocation points of presence you have, the bigger the presence, the faster it usually goes and the more you see. So the idea is if I'm at scale on my engineering side and my developers are optimizing the back-end engine or the architecture, and it's set up in a way for me to do that, I can do it faster.”

(As an aside, we note that Index Exchange, an SSP and ad marketplace much smaller than TTD, said it can burst to 10 million QPS on a recent Marketecture podcast[14], so maybe TTD’s QPS is not such a distinction.)

Finally, one consistent piece of feedback we heard from formers and corroborated by our perusal of ad tech forums is that TTD has built a reputation for phenomenal customer service that is second to none. Even though The Trade Desk is a self-service platform, the client success representatives are very hands on from training to campaign management and optimization, almost to the level of a managed service but without the associated cost to the client. This is another advantage that not only scales as TTD captures a greater share of its largest clients’ ad budgets, but is also prohibitively expensive for competitors to replicate as they lack the expense leverage to offer such services on an ongoing basis for free.

2) Independence

Since its founding, TTD has retained an exclusive demand side focus, with management regularly touting the virtues of being independent from the supply side (owning no SSPs or ad inventory) and from conflicted walled gardens. The reasoning is straightforward: ad buyers are the most important participants in the advertising ecosystem because they control the ad budgets, and in the world of digital media where ad inventory is effectively infinite, supply greatly exceeds demand so bargaining power rests in the hands of the buyers.

Management assert that by being the trusted independent DSP, clients trust the platform with their most granular and expressive 1^st party customer data that they would never share with the likes of Google or Amazon. This 1^st party data can then be leveraged alongside TTD’s 3^rd party data integrations to more effectively model and target the most relevant audiences. As Jeff Green explained during the Q2 2023 earnings call:

“As [advertisers] understand more about the power of programmatic, they are choosing the objectivity and performance of our platform over the murkiness of walled gardens more than ever before.”

“Advertisers and media companies are increasingly aware that they need to work with the technology partners who represent their interests, not conflicts of interests. As a result, advertising dollars are shifting to our platform and we will continue to invest in our platform and push the industry to create even more value for advertisers on the open internet.”

This alignment with the advertiser and lack of owned ad inventory also makes for good marketing PR. Within ad tech, the role of the DSP is to source the best inventory at the most efficient CPM for ad buyers, while an SSP’s role is to maximize the inventory yield for the publisher (ad seller). Naturally, these roles serve opposing interests, and any ad tech vendor that owns both demand- and supply-side solutions is subject to a conflict of interest.

At the extreme is Google’s ad tech business, which owns the two largest DSPs in Google Ads and DV360, the largest ad exchange AdX, and the largest SSP and ad server Google Ad Manager (formerly DoubleClick for Publishers). Owning the entire end-to-end stack at Google’s scale allows its DSPs to prioritize Google’s O&O inventory, its AdX to see substantially all auctions and bid prices across the open web before placing its own bids, and its ad server to award impressions to Google bids even if they are not the highest CPM bid.

However, TTD’s dogmatic adherence to independence does have a key downside: it does not own or control any supply and thus can’t tap into inventory at a lower cost for its clients. Part of the theoretical value proposition of an ad tech vendor owning both a DSP and an SSP is the ability to connect ad buyers with publishers at a reduced fee, thus allowing more of an advertiser’s media dollars to reach the pockets of the publisher (and all fees accruing to the single ad tech vendor).

TTD is also entirely dependent on its supply partners making their inventory available to it, though TTD brings significant advertiser demand such that SSPs and independent publishers are unlikely to withhold inventory from TTD if they want to optimize their yield. In fact, the company’s singular focus on the demand side is being increasingly diluted by its push into Supply Path Optimization, which we will discuss later in our analysis.

3) Close relationship with ad agencies

As mentioned in the history section, TTD made the early decision to align with the agency holdcos during a period when agencies were under threat of disintermediation by the rise of programmatic advertising. By getting the agencies onboard early and elevating their relevance in a programmatic world, TTD was able to turbocharge its growth because the agency holdcos controlled so many ad dollars. This agency alignment remains a differentiator for TTD and has helped the company survive competitors that elected to chase advertisers directly and alienated the agencies – e.g. close peer Rocket Fuel’s demise came in part due to trying to circumvent the agencies.

TTD’s agency relationships will be particularly advantageous during the migration of linear TV ad budgets to CTV. The agencies control the linear TV budgets of the largest global and national brands and are likely to continue to do so through the CTV migration. There are two main reasons we believe TTD is in an advantageous position:

1. Ad agencies are extremely siloed organizations with channel-based media buying teams controlling their own buckets of ad dollars and vying for greater share of their clients’ overall budgets. It is not yet clear whether the linear TV buyers or the programmatic buyers will emerge the primary custodians of CTV dollars, though TTD’s strong relationships at the senior agency and holdco levels should allow it to win regardless; and

2. Agencies are the “ad whisperers” to their brand clients and need to be convinced of the value of programmatic CTV before they shift their clients’ linear TV budgets into this emerging channel. TTD’s role as trusted partner to the agencies places it in prime position to accelerate the migration of ad budgets to CTV.

However, note that the status quo is changing. Proctor & Gamble, formerly the world’s largest advertiser by ad spend (dethroned by Amazon), began bringing its media activities in-house in 2018, which catalyzed a shift in how brands viewed their agency relationships. This coincided with a push by TTD to build more direct brand relationships through joint business plans (JBP) with large advertisers. Indeed, TTD is a beneficiary of P&G’s in-housing, with P&G’s growing programmatic CTV investments largely being handled by TTD.[15]

We understand that over the last three years, TTD has built out a large client development team that works directly with brand clients, making it easier for brands to control their ad tech stack, their valuable 1^st party data, and their media planning decision making process. Today, an estimated 20% of TTD’s gross spend comes directly from brands. Per CEO Jeff Green on the Q2 2022 earnings call:

“Joint business plans, or JBPs, are long-term deals that we sign with leading brands, who aim to increase spend on our platform, often over a multi-year period. In many cases, these agreements are signed with the partnership and cooperation of the agencies that represent the brands. So, as we continue to move upstream and get more direct commitments from CMOs and brand marketers, it not at the expense of our valuable agency relationships.”

We note that Green is careful to couch these JBPs in terms that are neutral to the agencies, but we do not believe this is always the case. One Tegus expert intimated that before TTD agrees to work with a brand directly, it would ask the brand’s (usually former) agency for permission so as to not antagonize the agency. We found this to have a whiff of marketing spin – imagine if your partner or spouse left you for another person, and six months later that person asks for your permission to marry your ex – does that make you feel better towards that person?

Almost every other source we heard from suggested that TTD’s push into direct brand relationships was fraught with risk and needed to be carefully balanced against the agency relationships. This is particularly the case for CTV, where TTD could accelerate the CTV migration by convincing advertisers directly, but this may alienate the agencies in their role as media planning advisors to their clients. As we discuss later in the CTV section, this is perhaps a risk worth taking, as there exists a counter-risk that agencies may be the ones disintermediating TTD in CTV.

4) Greater transparency and control

Management makes a big deal about TTD’s transparency and control – both in terms of decisioning and pricing. Giving control back to ad buyers appears to be a mantra of TTD along with its stance on independence. One of the most common advertiser criticisms of walled gardens is the loss of control – not only over the choice of ad tech (being forced to use the walled gardens’ tools), but also over ad buying decisioning itself. This criticism is becoming more prominent as AI-powered automated “black box” campaigns such as Google Performance Max and Meta Advantage+ wrest all low-level decision-making control away from the ad buyer. These black boxes are also very opaque when it comes to measurement, attribution, and reporting, often leaving advertisers in the dark regarding how their budgets were spent and how the ROAS was achieved.

Per TTD’s 2022 10-K:

“Our platform is transparent and shows our clients their costs of advertising inventory and data, our platform fee and detailed performance metrics on their advertising campaigns. Our clients directly access and execute campaigns on our platform and control all facets of inventory purchasing decisions. Clients also receive detailed, real-time reporting on all their advertising campaigns. By providing transparent information on our platform, we enable our clients to continually compare results and target their budgets toward the most effective advertising inventory, data providers and channels.”

Based on our channel checks, TTD is also one of the most transparent ad tech vendors when it comes to fees. On top of its base platform fee, TTD charges a la carte fees for each value-added feature or data integration that the client turns on. While the company is well known and sometimes criticized for nickel and diming its clients on just about everything on the platform, it at least offers the fee transparency for clients to adjust their use of premium features and data to optimize their TTD invoice against their campaign outcomes. Given some of the damning fee studies conducted in recent years and the general downward pressure on ad tech take rates which we discuss later in the report, this transparency does shield TTD’s take rate somewhat as it can more clearly point to where its incremental fees are adding value to a campaign.

Having said the above, we have seen TTD launch new tools that introduce a layer of opacity to both the decisioning process and fees. For example, Kokai, the next iteration of TTD’s Koa AI that will roll out through 2024, will insert more machine learning throughout the ad buying workflow. Users can turn on real-time ML-driven campaign optimization but this is likely to come at the expense of some level of control and agency. Management couch these optimization capabilities in terms of “recommendations” with the user retaining ultimate control over every ad buying decision, but it should be obvious that no human, especially not TTD’s core user – the recently graduated, overworked and underpaid agency trader handling a dozen campaigns simultaneously – can possibly respond in real time to the optimizations that are being “recommended” by a platform that views 13 million QPS.

Couple the above with the Koa bid shading tool where TTD also optimizes the CPM that is bid for each auction and takes a cut of the purported savings, and Kokai starts to look more and more like Google’s Performance Max or Meta’s Advantage+ campaigns with none of the benefits of their vast pools of 1^st party data or O&O inventory to optimize against.

In the remaining two-thirds of this report, for paid subscribers, we discuss our views on:

1. The CTV and retail media opportunities and why they might not be as exciting as management likes to project;

2. The sustainability of TTD’s take rate;

3. The potential impact of 3^rd party cookie deprecation; and

4. Some concluding thoughts on valuation.

Dissecting the growth drivers and other opportunities

One of the main attractions of TTD as a business is its overweight exposure to the fastest-growing channels in digital advertising: CTV and retail media. Management takes every chance to hype up the potential of these opportunities, so in the following sections we will assess each in turn with a rationally skeptical lens. We will also discuss opportunities and implications of Unified ID 2.0 (UID2), TTD’s alternative identifier for advertisers, and supply chain optimization initiatives around OpenPath, ultimately tying back to the key debate around the sustainability of TTD’s take rate.

“These dual forces of retail and CTV are truly bringing the power of data-driven decisioning home for advertisers and it is accelerating their shift to programmatic on our platform.” – Jeff Green, Q2 2023 earnings call.

Connected TV

Connected TV can be broadly defined as video content that is streamed over the internet on connected devices such as smart TVs, over-the-top devices, smartphones, and tablets. There are two revenue models for CTV:

1. Subscription video on demand (SVOD), where services charge a recurring subscription fee for access to ad-free video content; and

2. Ad-supported video on demand (AVOD), where ad-supported content is offered for free or at a lower subscription price.

The most prominent SVOD services are Netflix, Disney+ and Amazon Prime Video, and popular AVOD services include Hulu and Peacock. There is a third category of Free Ad-Supported TV (FAST) channels which represent the long tail of typically non-premium content that are available either on demand or linearly, but still fall under an ad-supported model as the name suggests. In the US alone, there are over 1,400 FAST channels amalgamated under providers such as Tubi, Pluto TV and the Roku Channel[16].  

For the purposes of this discussion, we focus on the US market as i) US data is most readily available; ii) the US CTV market is most developed; and iii) 87% of TTD’s gross spend in Q3 2023 was in the US. We also refer only to the market for professionally produced content and set aside user generated content platforms such as YouTube and Twitch, as these are not addressable by TTD. YouTube, including YouTube TV, is the largest CTV platform globally ex-China with $29.2 billion of ad revenue in 2022 compared to just $3.1 billion of estimated CTV gross spend on TTD as the largest independent DSP for CTV advertising. However, YouTube ad inventory is only accessible via Google Ads and DV360, Google’s DSP for the open web.

We observe a wide range of CTV ad market estimates, with GroupM being the most conservative and eMarketer having the most aggressive forecasts, though note GroupM excludes YouTube TV while eMarketer and Morgan Stanley both include YouTube TV. Global CTV advertising growth saw a huge spike during the pandemic, supported by the launch of several premium US AVOD services (Peacock, Discovery+, Paramount+ and HBO Max). GroupM’s forecasts look overly conservative considering the recent launch of ad-supported tiers by Netflix, Disney+ and Amazon Prime Video, especially as it also excludes YouTube TV. We see eMarketer and Morgan Stanley’s forecasts as a reasonable range for US CTV ad spend, which is the most important CTV market for TTD.

Source: GroupM (December 2023); eMarketer (October 2023); Morgan Stanley (December 2023); Bristlemoon Capital

Source: GroupM (December 2023); eMarketer (October 2023); Morgan Stanley (December 2023); Bristlemoon Capital

What has driven the rapid growth of CTV since the mid-2010s? The biggest factor contributing to the rise of streaming has been cord cutting by US households, concurrent with the rising adoption of Netflix, Hulu, Prime Video and more recently Disney+ and other network-owned streaming services. This report is a deep dive on TTD so we will spare the details of the decline of linear TV except to say that Netflix’s tenacity and the cable networks’ greed catalyzed a series of irreversible events that have not only reshaped the US media landscape but also the advertising landscape.

Source: Nielsen; Bristlemoon Capital

Source: Morgan Stanley (December 2023)

The decline in linear TV ratings means that broadcast and cable TV advertising no longer has the national reach it once did. Nielsen estimates that total US pay TV viewing hours has declined -65% from 2015 to 2022, compared to a -3% decline in linear TV ad spend according to eMarketer (note this includes both pay TV and broadcast TV). There is simply a shrinking and aging audience to activate multi-million-dollar national ad campaigns against, as 61% of the valuable 18- to 29-year-old cohort have never subscribed to cable or satellite at home before according to a 2021 Pew Research survey[17].

“I said incidentally at the last Investor Day, that traditional television and cable television is a ticking time bomb that we have mistakenly created like 3% declines and that will go on for 20 years. That’s not what’s going to happen. It’s going to get to a place and then fall off a cliff. And it does accelerate as it becomes less and less economically viable. We’re already at that point where it’s accelerated.” – Jeff Green, 2022 Investor Day

Yet linear ad budgets have been slow to migrate, given the decades-long relationships between ad buyers and publishers (TV networks), grandfathered preferential deals at lower-than-market CPMs, and unmeasured entertainment and other kickbacks. Multiple industry experts we heard from indicated shifting budgets from one channel to another was more complex than people realize and can take a long time. Firstly, existing contracts and relationships need to be untangled and sunset. Then, advertisers and their agencies need to test the programmatic channels and form new relationships before they can commit serious ad budgets. Agency linear TV buyers also face career risk– will they be replaced by the programmatic buyers? Fewer humans are required to manage programmatic buying compared to manual directly negotiated orders.

But urgency is building. The fact that linear TV ad spend has declined slower than viewership means CPMs have increased while reach and efficacy of linear TV advertising has declined. At the same time, the cable networks that tried to recreate Netflix’s streaming success have realized 1) producing content exclusively for streaming is expensive, 2) operating a streaming service is expensive, and 3) consumers are only willing to pay so much every month for entertainment. Premium streaming platforms including Netflix have all turned to ad-supported tiers to amortize their content costs over larger audiences and generate incremental ad revenue.

This influx of premium CTV ad inventory from the largest streaming services Netflix, Disney+ (both launched ad-supported tiers in late 2022) and Amazon Prime Video (January 2024) should help accelerate the shift of linear TV budgets to CTV. These services have huge audience reach (albeit likely with substantial subscriber overlap that will need to be managed by a capable DSP), highly engaging content, and are generally considered brand safe with minimal ad fraud compared to UGC platforms and FAST channels. Despite being relatively early days for premium ad-supported streaming, every streaming giant with both an ad-supported tier and an ad-free tier (including Disney, Netflix, Paramount, Warner Brothers Discovery and NBCUniversal) has said that total revenue per user is higher on the ad-supported tier than the ad-free tier[18].

Premium CTV services typically have ad loads in the 4-5 minutes per hour range compared to linear TV ad loads of 15-20 minutes in the US. This combination of attractive channel characteristics and limited ad supply should drive higher CPMs for premium CTV inventory. Indeed, we saw this with Netflix launching its ad-supported tier at $60 negotiated CPMs and Disney+ at $50 CPMs, though both have fallen over the past year as more inventory (growing ad-supported subscriber bases) and weak ad sales (for Netflix at least) have caused their initial premiums to contract. Amazon is expected to launch its ad-supported tier at a more competitive low-to-mid $30s CPM. Even so, these premium CTV CPMs should remain materially higher than average linear TV CPMs in the $10-20 range[19] as demand is likely to exceed supply for the foreseeable future.

Source: eMarketer (September 2023)

Source: OC&C Strategy Consultants

The other reason for higher CTV CPMs, and a key difference to the open web and linear TV, is that substantially all CTV viewers are logged into their service. That is, these are authenticated users with verified emails who can be identified by publishers and advertisers. This compares to just 5% of open internet website users who are typically logged in, according to TTD[20]. We will discuss this further in the Unified ID 2.0 section later, but a logged in user is an identifiable user, and an identifiable user can be targeted by advertisers.  

“And with CTV, brands can apply data to their TV campaigns and measure with much more precision, the effectiveness of every ad dollar. And much like retail, advertisers value the authenticated audiences that come with CTV – pretty much every streaming TV viewer is logged in with an email address.” – Jeff Green, Q2 2023 earnings call.

So far, so good – CTV looks to be a tremendous opportunity for TTD and one on which investors can hang a 56x 1-year forward P/E multiple on. In fact, don’t take our word for it; watch this short video of Jeff Green himself promoting TTD’s CTV opportunity during his recent appearance on CNBC’s Squawk Box, before Becky Quick calls him out on his antics.

Jeff Green makes two broad claims about the CTV market in his communications to investors: firstly, he claims the CTV market is perfectly fragmented, with the implicit benefits accruing to the largest demand platform in a fragmented supply environment; and secondly, that all CTV will eventually move to programmatic. Let us assess each of these claims in turn.

“Number two, I would say is, we have to win in CTV, and the good news is, in CTV, that the market around the world is perfectly fragmented.” – Q3 2021 earnings call.

“I don’t think any CTV or broadcast company can form a walled garden strategy in connected television. The landscape is nearly perfectly fragmented where nobody has dominant market share in television.” – 2022 Investor Day.

“As I’ve said before, I believe television is perfectly fragmented for a business like ours.” – Q3 2023 earnings call.

“But it does serve our belief that at end-state all things will be digital and all things digital will be programmatic.” – 2022 Investor Day.

“Every premium video content company from Disney to Paramount to NBCU and Sky, to Netflix have changed pricing and embraced advertising. Not all of them have yet embraced or centered around programmatic advertising, but we predict they will.” – Q3 2023 earnings call.

Is the CTV market perfectly fragmented?

To answer this question, we need to define what “perfectly fragmented” means. Is it “perfectly” fragmented in the economic sense, like “perfect” competition? Clearly not, given the millions of websites on the open internet compared to the dozens of CTV services that generate any kind of meaningful viewership. But is it perfectly fragmented as in just the right amount of fragmentation for a large ad demand platform to take advantage of? That’s the narrative that Jeff Green wants to spin: because the CTV market is not so fragmented as to be unscalable to serve, but also not so concentrated that any one or few services have power, all these large but not dominant publishers need to depend on the advertiser demand that TTD brings.

“This will happen in part because CTV is perfectly fragmented, but collectively huge. It is not so fragmented that you need millions of parties to coordinate, but it’s fragmented enough that no one has enough power to be draconian and do it alone.” – Jeff Green, Q4 2022 earnings call.

We believe this logic may have been sound when TTD first launched its CTV tools in 2016, right up until late 2022 when Disney and Netflix launched their ad-supported tiers. Prior to this, we believe TTD may have been mainly buying impressions on what could be considered “second rate” premium CTV services: the Hulus, Peacocks and Pluto TVs of the world. In fact, up until COVID, TTD may have been largely buying ads on Roku TV, Hulu and FAST channels – we suspect this because the FreeWheel H2 2020 U.S. Video Marketplace Report[21] showed streaming services only had 38% of digital video ad views, and the FreeWheel H1 2022[22] report showed FAST channels still had 29% of ad views and DTC only 13% (both fall under streaming services).

These are mostly publishers who have limited to no capability or incentive to build their own ad tech. In this world, TTD may have gained the impression that its buying power was necessary for these CTV services to maximize their ad yields. But now that the 800lb streaming gorillas have entered the AVOD market, we cannot be so sure that the future will resemble the past.

Source: Redburn

The way we see it, premium CTV is not perfectly fragmented. In fact, we would argue that it is quite concentrated in the hands of a few big publishers. Looking at the Nielsen share of streaming viewing data below, we note the following:

• YouTube is a quarter of the CTV market and off limits to TTD (inventory only accessible through Google’s DSPs).

• Netflix is the second largest CTV service and is partnered with Microsoft’s Xandr as its exclusive DSP. Yet barely nine months into their relationship, in March 2023, Netflix was reported to be reviewing its ad strategy and considering a build-or-buy pivot away from Xandr[23]. We note that Xandr provides the full ad tech stack to Netflix: ad server, SSP and DSP. It could be that Netflix ultimately decides to in-house the ad server and SSP components (the most important pieces for deciding which ads to run against which viewers) and open up its ad inventory to other 3^rd party DSPs, which would include TTD. Alternatively, Netflix could in-house the entire stack and likely will have the scale to force advertisers to buy through an in-house DSP. Even if Netflix does open its ad inventory up to 3^rd party DSPs, we would not preclude them from changing their mind and in-housing it later (remember when Netflix said it would never run ads?). There is also precedent of this happening, e.g., when Facebook moved its 1^st party data behind its own DSP and effectively killed the 3^rd party DSPs buying on the Facebook Exchange back in 2016.

• Hulu inventory remains open to 3^rd party DSPs, though we understand that some top-flight inventory may be reserved for its own ad buying tools[24]. However, Comcast’s 33% stake is about to be acquired by Disney, who will then own 100% of Hulu. Disney has its own ad server and ad exchange (SSP) that is used for Disney+. While Disney+ inventory is currently available to 30 DSP partners including TTD, we would once again not rule out Disney combining the ad tech stacks of Hulu and Disney+ together and in-housing the DSP function once it has built sufficient scale, data connections and targeting capabilities.

• Amazon Prime Video is moving to ad-supported in January 2024 and inventory will be available through Amazon DSP exclusively. It seems unlikely that Amazon would open up Prime Video to 3^rd party DSPs and share its incredibly valuable 1^st party shopper data.

• Disney+, as mentioned above, has its own supply-side ad tech already. It has also integrated Hulu’s DSP, which it has renamed Disney Campaign Manager. It has all the pieces to in-house the full stack should it choose to. Even if inventory remains open to 3^rd party DSPs, we believe Disney will aim to minimize ad tech fees by courting direct relationships with brands (which it already has through 100 years of linear TV advertising).

  • Update: Disney, Fox and Warner Bros Discovery announced last week (February 7) a pioneering sports streaming service that will bundle the sports content of all three networks in one app. The JV will be owned in 1/3^rd shares, but let’s be honest – it will be the Disney and friends service. If any content were to entice advertisers to integrate with another DSP, it is sports content. We believe this service may further incentivize Disney to in-house the full ad tech stack, especially if Disney is also responsible for handling ad sales on behalf of Fox and WBD.

• Apple TV+, which is not broken out in the data, is likely to in-house its ad tech if it ever launches an ad-supported tier given the company’s philosophy around control and privacy.

• Max, Peacock, Pluto TV, Tubi, Roku TV and the long tail of streaming services and FAST channels are unlikely to ever achieve the scale to fully in-house the DSP function, though could potentially reserve some of their “premium” inventory for in-house buying tools. Roku, after acquiring Dataxu (a mid-market DSP), tried to build a quasi-walled garden by fencing its 1^st party data behind its in-house DSP. The company eventually had to reverse course after it realized that unlike Facebook, its 1^st party data did not have the gravity to pull ad buyers away from the industry leading DSPs such as TTD and DV360.

Source: Nielsen; Bristlemoon Capital

If we sum up the above, there’s:

• ~55% of CTV viewership (YouTube, Netflix and Amazon) not available to TTD, noting that Netflix could decide to open its inventory to other DSPs, or in-house it after the Xandr partnership expires;

• ~17% that could eventually be in-housed (Disney+ and Hulu);

• Leaving less than 30% of CTV viewership open to TTD (and other 3^rd party DSPs) with minimal risk of in-housing.

This doesn’t seem like a perfectly fragmented market to us.

We see two disintermediation risks that TTD faces during the linear to CTV migration:

1. Walled gardening

2. Agencies going direct to publishers

We already discussed which streaming services have the scale and/or valuable 1^st party data to potentially erect walled gardens, so won’t belabor the point again. We would just add that advertisers and agencies are increasingly consolidating their spend through a single or small handful of DSPs to better track and measure their ad spend across channels and publishers. CTV services will need to carefully balance their desire to minimize or eliminate ad tech fees via in-housing the stack against achieving sufficient auction density to optimize the yield on their inventory. We have not seen any studies analyzing the impact of auction density on yield optimization and would hesitate to speculate on where the drop-off point might be.

On the risk of agencies going direct to publishers, this is likely underappreciated by many. Jeff Green and team spin a compelling narrative about the mutual importance of TTD and the agencies to one another. However, as we discussed earlier, TTD has been expanding its advertiser direct relationships, and it should not be lost on anyone that agencies are doing the same in the opposite direction. Agencies are under increasing pressure from their clients to improve media budget efficacy and minimize fees – basically to justify their existence – and DSPs being the next link in the supply chain are the natural vendors to squeeze.

Agencies are well accustomed to working directly with the handful of national broadcast and cable TV networks that matter in the old world of TV advertising. There is no reason to believe they wouldn’t be equally capable of working with the three or four most important CTV services on direct deals and PG or even PMP campaigns. If the agencies can squeeze or cut out the DSPs’ fees and protect their own thin fees, they will be incentivized to do so.

We believe this is already happening. The most recent development is Publicis’ launch of its CoreAI platform, which has DSP-like functionality and leverages Publicis’ Epsilon 1^st party data and ad tech solutions. We learned that during an analyst Q&A, management purportedly confirmed that Publicis absolutely intends to move into the DSP space and squeeze fees out of the channel. Days later, WPP announced its own big bet on AI. While we are not sure if WPP harbors similar ambitions to encroach upon the DSP space, it seems likely the large agency holdcos would follow in Publicis’ footsteps given the highly competitive nature of the industry.

We would also speculate that while US AVOD and FAST ad spend was in the sub-$20 billion range[25] – i.e. before Disney+, Netflix and Amazon’s ad-supported tiers were launched – advertisers and agencies may have been accepting of 3-15% DSP fees (we discuss fees in the next section). But as more and more of the remaining $60 billion-plus linear TV ad spend shifts to CTV – remember, Jeff Green believes linear TV will fall off a cliff – DSP fees start to add up to significant sums. CTV buyers might balk at this fee, particularly as linear TV buying has zero ad tech tax.

There is actually a third disintermediation risk (and opportunity) in the form of supply chain optimization, which we will discuss later in its own section.

What does “shift to programmatic” actually mean?

Even if CTV is perfectly fragmented, what does Jeff Green actually mean when he predicts that all CTV ad buying will shift to programmatic?

We learned in the introduction to programmatic advertising section that there are three main programmatic buying models: Programmatic Guaranteed (we’ll lump preferred deals in here for brevity), Private Marketplaces, and open exchange auctions. While all three are programmatic, they can come at very different CPMs and fees.

Firstly, let’s establish what percentage of CTV ad buying is currently not programmatic. Data on the share of direct deals in CTV advertising is mixed: FreeWheel, a CTV ad tech platform owned by Comcast, reports that 65% of ad views on its platform in H1 2023 were non-programmatic direct, down from 76% in H1 2021 (i.e., programmatic has been gaining share)[26]; Digiday estimates that direct orders are as high as 70% of CTV ad sales[27]; and TTD showed a chart at its 2022 Investor Day that pinned direct orders at 56% of CTV ad sales, PG at 19% and “decisioned programmatic” at 25%.

Source: The Trade Desk; Bristlemoon Capital

We did not come across any time series data for programmatic share of CTV aside from FreeWheel’s semi-annual Video Marketplace Reports, which do show that programmatic buying has been gaining share over recent years. Unfortunately, the reports do not split the programmatic category further into PG and programmatic biddable (aka decisioned programmatic) which includes PMP and open exchange. But we can reasonably conclude that CTV advertising will continue to shift towards programmatic over time. For example, Disney+ has said all of its inventory will eventually shift to biddable[28]. We suspect “biddable” here means PMP, as it is very unlikely a premium publisher like Disney courting premium brands will release inventory into open exchanges.

However, we would note that one ad tech expert we heard from had the following to say about the shift to programmatic:

"There’s some resistance. Even some of the media owners, like Fox and Disney, externally, they say all the right things, “We’re all in on programmatic, we make all our inventory available with programmatic, us and our amazing content, plus The Trade Desk’s technology and data is the best of both worlds.” However, they’re going to drag their feet as long as they can, the networks that is, and even probably the Netflixes of the world, to continue to bring in dollars the old-fashioned way, meaning, “Send us a contract or an insertion order, we’ll retain control over how that inventory is parsed out, targeted, and how we manage our margins.” – Former GM, Business Development at The Trade Desk.

This is corroborated by an analyst question on the Q3 2023 earnings call.

Analyst: “When I talked to Disney and Paramount and Fox, they are always saying they have direct sales forces, they think they get a higher price point selling content, which is not substitutable, and they're definitely afraid of this third bucket that you talk about, which is what you think is the final destination, which is biddable programmatic, because they've seen what's happened in mobile and desktop, and it's sort of a race to the bottom because of unlimited supply.”

Jeff Green: “Yeah, so, I suppose it depends on who you ask inside of those premium content owners, but to me, the math is fairly obvious, and even conceptually, it's fairly obvious.”

We don’t think it’s fairly obvious, so let us explain.

Programmatic fees

You might ask, “why does all this matter, which bucket the ad spend falls in”?

One word: fees.

Through our research, we understand that direct orders might earn a DSP a 1% fee at best, to the extent they participate at all. This is because everything is negotiated between the publisher and the ad buyer, with any DSP involved simply acting as the conduit to pass ad dollars from buyer to publisher. PG deals have a fee of 3-5% for the DSP – there is no bidding or decisioning taking place, but the DSP’s tools are used to automate the ad buying process, so they earn a nominal fee.

Once we get to programmatic biddable, this is where things get murkier. The lowest fee range we heard from one sellside analyst was 5-8% for PMP based on their industry conversations (not specific to TTD), which we could not corroborate with anyone else. The analyst also pointed to an SSP blog[29] that suggests Preferred Deals are <10% DSP fee and PMPs ~10%, with open auctions at a 10-20% fee, though this is not specifically for CTV ad buying. Others we spoke to either could not confirm a range or thought that PMP and open exchange auctions had similar fees without substantiating evidence. One Tegus expert said TTD’s platform fee is 10-15% of media spend (with data fees layered on top) and did not suggest there was any meaningful difference in platform fee across channels. But he also noted that the platform fee for large agency clients could be less than 6% given their buying volumes.

To the extent that CTV ad buying shifts to “programmatic”, it doesn’t tell us a whole lot about the economics for TTD. Redburn, one of only three Sell rated analysts out of 33 who cover TTD, does not seem to distinguish between PMP and open auction within programmatic biddable, or distinguish between CTV and other channels for that matter when they come to assessing take rates. They note that going forward, CTV take rates will be substantially lower than the current ~20% company-wide take rate, with 10% or less being plausible. This is due to a mix shift toward larger brands and more concentrated agency spend in the CTV channel relative to the open web, and a shift to more direct deals and PG on the platform where TTD’s real time bidding and optimization functions are less important.

We will put a pin in this topic of CTV fees for now and return to it when we discuss the sustainability of TTD’s take rate in a later section.

Programmatic CPMs

The other reason Jeff Green is pushing the shift to biddable narrative is that the company believes biddable CPMs will be higher than negotiated CPMs. This seems logical – the more data an advertiser can apply to bid on specific audiences, the more valuable those audience impressions are and the higher the CPM. As guaranteed impressions do not leverage data nor price discovery, they should in theory be priced at lower CPMs than biddable impressions.

“So to put a finer point on this, auction-based selling for CTV is the way of the future. There is no way that Netflix will be able to command the CPMs that they're talking about right now in marketplace in the future without an auction and without an underlying framework of identity to enable buyers to activate their data…And again, the only way that Netflix is going to be able to command the CPMs that they hope to get to continue to monetize their content is through an auction model just like that because that's where those CPMs live. They live in a biddable auction-based environment.” – Chief Revenue Officer Tim Sims, 2022 Investor Day.

To support this claim, Mr Sims showed a chart in his Investor Day presentation plotting CTV bid prices (y axis) against bid count (x axis), based on one day of observing CTV bids on the TTD platform. This chart suggests that there are a huge number of bids on programmatic biddable inventory that are made well above where direct and PG deals typically price. The implication being that CTV publishers are leaving vast amounts of money on the table by favoring negotiated CPMs over price discovery in an open auction.

Source: The Trade Desk 2022 Investor Day

We would make the following observations:

1. The 2022 Investor Day was held on October 4^th, 2022, just before the Disney+ and Netflix ad-supported tiers were launched. We know Disney+ and Netflix launched with $50-$60 negotiated CPMs, and while these have fallen to the mid-$40s, they are still much higher than the PG pricing band of $18-$23 shown in the chart. That band represents the average PG campaign on the TTD platform.

2. The x axis is conveniently in log scale. What looks like a huge number of bids in the $40+ CPM range probably only adds up to one or two blue dots in the $20 CPM range. Even based on this chart alone, the majority of bids appear to be less than the negotiated CPMs that Netflix, Disney+ or even Peacock receives.

3. The chart conveniently ends just past 100m on the x axis. In the same presentation, there is another slide stating that TTD sees 30 billion+ CTV impressions per day. We get that the platform doesn’t bid on 100% of the impressions it receives, but it seems to us the dots on the chart do not sum to even 10% of those 30 billion impressions. If management had shown this data as a frequency distribution instead of a log scale scatter plot that ends just as it reaches the PG pricing band – which is what an intellectually honest statistician would do – we suspect that the distribution would be clustered around the $10-20 CPM range. This obviously doesn’t tell the story that management want the data to tell, so they have concocted in what our view is a favorably massaged chart that gels with their narrative.

4. Finally, we have no idea what kind of inventory was being bid on. The very small number of $100+ CPM bids could have been placed on highly niche CTV channels that reach very specific audiences that are valuable to certain advertisers. It is hard to extrapolate the data as presented in this chart to conclude that all CTV advertising will move to biddable. Again, a better study would have been to pick a single (preferably large) CTV publisher and observe the bids in that single environment.

Overall, we haven’t found compelling evidence that CTV CPMs are, as a rule of thumb, higher for biddable inventory than guaranteed inventory. Being able to leverage data to target specific audiences is powerful, but a real-time auction also allows transparent price discovery. Do premium CTV services want this kind of price discovery for all their inventory, or would they rather guarantee their inventory is sold at a fixed price, even if it risks leaving some dollars on the table?

This is not even considering impression “quality” at a single premium CTV publisher. Remember that “premium” here simply means the audiences are logged in and identifiable. But one Disney+ viewer might match to much more 3^rd party data than another Disney+ viewer and therefore be a higher quality impression, despite both being logged in with email addresses. Consider a high-quality impression that attracts a 2x higher bid compared to a negotiated CPM, and a low-quality impression attracts only a 0.3x bid. Whether the publisher’s overall CPM is higher under a biddable or guaranteed model is entirely dependent on the distribution of impression quality. We can’t simply say that biddable CPMs in aggregate are higher than negotiated CPMs.

As a case in point, Hulu is the premium CTV service that has been in the programmatic advertising game for the longest, having been ad-supported since 2016. Yet its CPM in the mid-$20s is only slightly higher than that of linear TV. See the CPM charts towards the start of the CTV section.

Will CTV be 100% biddable?

To be fair to Mr. Green, when he talks about the shift to programmatic, he is in fact talking about programmatic biddable while acknowledging there is a journey from direct to programmatic guaranteed first, before progressing to biddable. But the question remains – will all CTV advertising eventually shift to biddable?

“So this is kind of the typical life cycle of somebody launching a streaming app, which is that streaming app will launch, sometimes they'll do some insertion orders or category exclusives, then they'll slowly move into programmatic, sometimes with PG and then ultimately move to biddable. So I think this is pretty typical.” – Jeff Green, 2022 Investor Day.

We find it improbable that all CTV advertising will shift to programmatic biddable – programmatic, yes; but biddable, no. If CTV ad dollars were mainly drawn out of existing search or social budgets, which are highly targeted lower-funnel customer acquisition channels, then it is likely these dollars will seek similar targeted and decisioned inventory. But CTV ad dollars are mainly coming out of linear TV budgets. We need to remember that linear TV is the channel for massive audience reach and large brand campaigns where the goal is raising brand awareness and brand lift. There are simply not enough impressions – let alone targeted impressions – in the CTV channel for many large advertisers to run large brand campaigns against. Streaming now accounts for more than half of US TV viewing time[30], yet less than 10% of TV advertising inventory.

Source: Nielsen; Bristlemoon Capital

As an example, take this answer from Jeff Green on the Q2 2023 call in response to a question on guaranteed vs decisioned CTV:

“So we've recently done some studies on decisioning across the platform and compared it to PG, and we found that the value of decisioning or the performance of the decisioning on a PPA (Price Per Action) basis is 5x better in general, or if you were to say it in terms of savings, it's an 83% savings.”

Let’s take a step back and think about what this means. In order to have a 5x better PPA (i.e. 80% lower), the conversion rate must be 5x higher for decisioned than PG – i.e. the ads are targeted at audiences that on average are 5x more likely to click or whatever the action is. This assumes the CPM stays constant, when in theory it should increase because of better targeting.

For a finite audience pool with a constant percentage of audience that is interested in the advertiser’s product, e.g., 1 million and 1% who are interested, the only way to achieve a 5x higher conversion rate is to target 1/5^th of the audience and hope you’ve captured all those who are interested. If the CPM doubles because of targeting, then a 5x better PPA would require limiting ad impressions to 1/10^th of the audience. This is a stylized example but serves to make a point: in order to achieve higher ROAS given a finite audience population, an advertiser must target its ads to a smaller segment of said audience.

Source: Bristlemoon Capital. This stylized example assumes 1,000,000 total audience population; in real life this might be the 125 million US TV households. If the advertiser wants to reach a wider audience, its campaign conversion rate must converge upon the population conversion rate.

Note that the ad budget savings don’t necessarily get allocated to more campaigns. If an advertiser achieves its campaign goal using only 20% of the allocated budget because of better decisioning, some portion of those savings likely fall to the bottom line as incremental margin. Biddable CTV with the level of targeting that TTD envisions clearly moves TV advertising down the funnel and is not suitable for top of funnel brand building. Yet on the very next earnings call (Q3 2023), Mr Green makes an oxymoronic statement, claiming that CTV will be both the best top of funnel channel and also deliver the most effective (i.e., decisioned or targeted) advertising at the same time.

“And let's not forget that television has always focused on being a top of the funnel activity where you're creating awareness. I would argue that's where the heavy lifting of advertising is done, which is the process of winning hearts and minds. Hearts and minds are not historically won in 240 blue and black characters, they are won with moving picture and audio and emotion. And that work continues to be done at the top of the funnel in CTV. So I would argue that that CTV right now is the most effective television advertising that television has ever seen. And arguably, it's the most effective advertising at scale ever.” – Jeff Green, Q3 2023 earnings call.

Linear TV advertising is largely a top of funnel exercise for raising brand awareness, and much of those ~$60 billion ad dollars will need to find a similar top of funnel channel to deploy into. This is where PG deals in CTV come in – the advertiser can receive guaranteed impressions in line with the reach they want to achieve, and can set the high-level audiences to show ads to (e.g., by demographics, geography, time of day, genre of show etc). But even guaranteed deals are not enough – we’ve demonstrated above that there’s simply not enough CTV inventory to soak up $60 billion of linear budgets, so some of those dollars will inevitably leave the industry (i.e. the linear to CTV shift won’t be 1-for-1). We believe that for TTD to continue capturing outsized share of linear TV advertising as it migrates to CTV, the company must start to embrace the PG model, if not direct deals.

TTD is accustomed to operating in channels where supply vastly exceeds demand, where it and its clients can call the shots. Given the stark imbalance between CTV’s share of viewing time (50%-plus) and its share of ad inventory (sub-10%), we expect demand to continue to exceed supply, leaving the balance of power in the premium CTV publishers’ favor. Bill Wise, the CEO of MediaOcean, noted on a recent Marketecture podcast[31] that a single media buyer had asked to reserve 100% of Netflix’s ad inventory for two years. Moves like this suggest there is a large subset of premium advertisers who are in no rush to shift their CTV budgets over to programmatic biddable.  

What is TTD’s market share?

We wrap up our analysis of the CTV opportunity with a final word on market share. Based on management’s qualitative commentary, video is currently around 45% of TTD’s gross ad spend, with the vast majority of video being CTV. This suggests that CTV is likely to be at least 40% of TTD’s gross ad spend, or $3.1 billion in 2022. If we take GroupM’s estimate of 2022 global CTV ad spend of $26.3 billion excluding YouTube TV, this suggests TTD has a market share of 12%. Plenty of runway to continue outgrowing the market.

However, Redburn believe TTD has more like 70% market share of the relevant programmatic biddable market. This is based on Jeff Green talking to having the “lion’s share” of the biddable CTV market, and the 25% of the 2022 CTV market that is decisioned programmatic per the pie chart from the 2022 Investor Day. Redburn also assumes that all of TTD’s CTV spend is decisioned programmatic (PMP or open auction) and it doesn’t touch any direct or PG spend that clients might want to run through the platform. We find this last point difficult to accept – if an agency or direct brand client wants to run a PG campaign on Disney+, is TTD going to decline and tell them to find another DSP? That seems like a surefire way to upsetting and churning your client base.

Now we don’t purport to know what “lion’s share” here means – it could be 40% of the market if #2 is 10%, or it could be 70% of the market. If it is indeed closer to 70% than 40%, then TTD has a lot less runway than headline industry CTV spend would suggest. It means that TTD’s future CTV growth would be closer to the growth of programmatic biddable within CTV, and even that assumes TTD continues to take very high incremental share of the ad dollars that shift from direct and PG to biddable.

As TTD is widely considered the best CTV ad buying platform in the market, maintaining a high incremental share of programmatic biddable is possible. But the future growth rate of programmatic biddable within CTV is potentially a lot lower than TTD’s historical CTV growth rates (when it started at 0% share), especially as the premium inventory from Netflix, Disney+ and Amazon are still mostly being sold direct or via PG. The alternative scenario is TTD starts handling more direct and PG ad dollars which would allow it to maintain higher growth rates but puts pressure on take rates.

Retail media

If the CTV opportunity seems somewhat inflated, surely the retail media opportunity can plug the gaps? After all, it is a $500 billion TAM according to the CEO:

“Shopper marketing [aka retail media] is, of course, when you add loyalty programs and sponsored listings and all those things, you get to roughly a $500 billion TAM. And I think over time, most of that is available to The Trade Desk.” – Jeff Green, 2022 Investor Day.

Retail media is a form of digital advertising that targets consumers at or close to the point of purchase, typically on retailer websites and e-commerce marketplaces. It is not an advertising channel per se as ads can appear on the web, in mobile apps, on CTV, and even in-store on digital displays. Retail media is a powerful avenue for advertisers because it can leverage a retailer’s rich 1^st party customer purchase data to target specific audiences.

Source: GroupM (December 2023); Bristlemoon Capital

GroupM estimates global retail media to be $119 billion in 2023, of which $41 billion is in the US and the vast majority of the remainder in China (Alibaba, JD.com, Pinduoduo etc. all run advertising on their marketplaces). Retail media as it is understood today has largely been driven by the success of Amazon’s advertising business, which is the largest retail media network ex-China. Amazon just recently reported $46.9 billion advertising revenue for 2023, which has grown at 39% CAGR since 2018. Retailers need only take one look at Amazon’s advertising growth, and one look at Meta’s advertising margins, to be sold on the promise of retail media. Since the 2017-18 US retailpocalypse, and especially since the start of the pandemic, every retailer with a website wanted to get into retail media.

Source: eMarketer (March 2023); Bristlemoon Capital

Onsite vs offsite retail media

Investors who only read or listen to what TTD says might not appreciate this but there’s actually two categories of retail media: onsite and offsite. Onsite retail media is advertising that takes place on a retailer’s O&O properties – e.g., Amazon.com or Walmart.com. When a shopper searches Amazon for a pair of running shoes and sees four sponsored products and a brand banner ad at the top of the results page, that is retail media at work.

Offsite retail media, on the other hand, is when advertisers leverage a retailer’s 1^st party data and their own customer data to target audiences on 3^rd party websites. These ads will link back to the retailer’s product page, or to the advertiser’s own landing page where the viewer can learn more about the product and complete a purchase.

According to eMarketer, onsite retail media makes up the vast majority of retail media spend, with offsite expected to gain 100-150 bps of share each year.

Source: eMarketer (March 2023); Bristlemoon Capital

Criteo and CitrusAds (owned by Publicis) are the two leading ad tech vendors for retail media. Onsite retail media requires three components:

1. A software solution that links to the retailer’s SKU-level data to know which products can actually appear as an ad;

2. SSP functionality that manages bid requests, the ad auction, and the serving of the ad on the retailer webpage; and

3. DSP functionality that allows advertisers to run automated campaigns on the retailer’s website and manages access to the retailer’s 1^st party data.

As we know, TTD only has the DSP functionality, and is very unlikely to build or buy the SSP capabilities given the company’s intention to serve the demand side exclusively. This means TTD is relegated to serving offsite retail media campaigns only, which as of today is only a small percentage of the total retail media landscape. TTD clients can also bid into onsite retail media auctions run by SSPs such as Criteo like they would for CTV or mobile display auctions.

However, it is not clear how this works in practice. Take Walmart for example – one analyst we heard from was told by Criteo that TTD is unable to buy Walmart onsite inventory (Criteo is Walmart’s onsite ad tech partner) and was told by TTD that it can buy Walmart onsite inventory. We, and the analyst, suspect this might boil down to access to Walmart’s 1^st party shopper data – it could be that the Criteo DSP has exclusive or privileged access to Walmart’s shopper data for onsite targeting, while other DSPs such as TTD have limited or no access. Without access to retailer 1^st party data, the appeal of retail media to advertisers is greatly reduced and becomes not much different to contextual targeting. Note this may not be the case with all onsite retail media, as some retailers with less data gravity than Walmart may share/sell their 1^st party data with all DSPs that buy on their websites.

We were initially concerned by TTD’s inability to fully address the onsite retail media opportunity given its lack of SSP functionality and potentially limited access to retailer data. However, one Tegus interview with a former Trade Desk and Criteo executive suggested there is limited money to be made on the SSP side of retail media given the size of the retailers being served.

“The problem with [onsite] retail media is there's no money in it. Again, it's just a tech fee.”

“Criteo will always be probably the majority player for the SSP portion of retail media. But other than that, that's really it. And again, there's no money in that. I know that for a fact.”

The money, according to the expert, is made on the 1^st party data, which the DSP makes available to advertisers for a fee and takes a cut of said fee. The expert estimates TTD’s share of the fee is 25-50%, while Redburn believes TTD’s data take rate more broadly is as high as 85%. It is not clear if the 25-50% take rate is in the context of onsite or offsite retail media. It feels high for onsite considering the retailer is sharing its own data to target ads on its own properties, though it could come down to the size and bargaining power of the retailer vis-à-vis TTD.

In-housing is already happening

Large retailers appear to be pushing back on DSP fees by in-housing the buying functions and data management. Criteo has flagged pressure on its Commerce Media Platform (i.e., retail media) fees on its Q3 2023 earnings call, disclosing that its largest client has decided to manage their largest brand advertiser relationships in-house, effectively cutting off Criteo’s DSP fee.

“What has changed for 2024 is our largest customer, we've recently engaged in a renewal, and we've moved to an economic pricing model that is more SaaS-like [instead of take-rate based]. That's a scale play, our mature play, we continue to drive massive value to them and we continue to also deliver, of course, value for us…And our expectation is we'll take, I would say, a shorter-term hit on the growth rate versus where we expect it to be in 2024.” – CFO Sarah Glickman, Q3 2023 earnings call.

“We've also seen already a shift to large brands being sold by the client [i.e., retailer] versus by Criteo. And our Commerce Media Platform capability being self-service already leans more into a platform pricing model.” – Sarah Glickman, Q3 2023 earnings call.

This should not come as a surprise. After all, retailers and brands tend to have close relationships, often stretching back decades. While a large retailer might lean on an ad tech partner to get its retail media business off the ground, once it is up and running it is not obvious to us why the retailer would continue to let ad tech extract its often-exorbitant fees. Large retailers are more than capable of handling their own ad demand sourcing from their brand suppliers and should only be willing to accept incremental DSP fees if the DSP can bring incremental demand.

The direct relationship between retailer and brand can give the retailer a better idea of the brand’s onsite media budget, unlike on the open web or the long tail of FAST CTV where ad buyer and publisher are anonymous and separated by 3-4 layers of ad tech middlemen. By comparing what a brand ought to be spending on media with what is ultimately received, the retailer has a good sense of where fees are leaking out and who to squeeze.

“The purchase of ad spaces or sponsored ad placements happens through the relationship teams and the shopper marketing teams on the brand side and the merchant and the retail media team on the retailer side. Agencies (edit: and by extension, DSPs), for the most part, and of course, everybody is structuring their teams a little different, don't play a strong role in that relationship.” – VP of Retail Media at Bed, Bath & Beyond Inc

TTD’s offsite retail media partnerships

TTD made its big splash on the retail media scene when it partnered with Walmart to be the tech provider for Walmart DSP, the retailer’s offsite retail media buying platform. TTD essentially built a white labelled instance of its main DSP platform for Walmart, to be used exclusively for buying ads on 3^rd party properties leveraging Walmart’s 1^st party shopper data. Any advertiser that wants to buy offsite retail media using Walmart data must do so through Walmart DSP and not the main TTD platform.

Despite the seeming importance of the Walmart partnership to TTD’s retail media narrative, we have been unable to ascertain even qualitatively how the economics of the partnership work out. Thus far, no one has been able to confirm who (Walmart or TTD) earns what fee when an advertiser activates on Walmart DSP or how data fees might be shared. Given the size of Walmart relative to TTD, our guess is that the economics are probably not favorable for TTD and the partnership is largely a marketing/halo exercise, but this is just a guess.

One TTD former in an October 2022 interview divulged that the Walmart partnership hadn’t been as lucrative as management expected, but quickly declined to speculate when pressed on why that might be the case. The Walmart DSP was launched in August 2021, so it could have just been a matter of nascency at the time of the interview rather than anything more sinister. But even one of the most ardent sellside TTD bulls opined that Jeff Green may have talked up the Walmart relationship into something bigger than it actually is.

Since the Walmart partnership, TTD has announced numerous other retailer partnerships including the likes of Walgreens, Kroger, Albertsons and Schwarz Group (Lidl and Kaufland). We understand these to be data partnerships rather than full platform integrations in the vein of Walmart DSP. We also believe that these are not real-time data integrations with retailers’ CRMs, but rather historical shopper data that is made available to advertisers to help them build campaign audiences to target. This is corroborated by how Jeff Green describes these data partnerships. We also do not know how many of these partnerships are exclusive, though we believe retailers have an incentive to limit the number of DSPs that have access to their highly proprietary 1^st party shopper data.

“With retail media, advertisers can start with precise authenticated data, whether that data is based on purchase history or loyalty data. From there, the advertiser can target based on what they know, and then do more accurate data-driven modelling to expand targeting segments with greater precision.” – Jeff Green, Q2 2023 earnings call.

We note that offsite retail media is typically used for retargeting campaigns – that is, targeting audiences that have previously engaged with the advertiser in some way – or for lookalike campaigns that target audiences with similar attributes to those who have previously engaged with the advertiser. Because offsite retail media crosses the open web, much of the targeting depends on 3^rd party cookies and measurement/attribution is also more difficult than onsite retail media, especially when factoring in the possibility of offline conversions.

How big is the (offsite) retail media opportunity?

In the fullness of time, offsite retail media spend should theoretically be higher than onsite retail media spend. Retailers have limited ad inventory on their O&O properties before ads start to negatively impact the shopping experience, compared to the near-infinite offsite 3^rd party inventory. We are not aware of any retailer that has reached an onsite ad load limit yet. Even Amazon, which has been doing retail media for over ten years, still appears to grow its predominantly onsite advertising revenues well ahead of its GMV growth.

Source: Morgan Stanley; Bristlemoon Capital estimates

We can say with all the confidence of one (rather unique) sample that a ratio of 7% onsite ad revenue to retail sales is not the limit of ad loads for onsite retail media. We would note that in China, Alibaba’s ad revenue reached just over 4% of platform GMV before growth abruptly stopped, though that is predominantly due to competitive pressures in Chinese e-commerce rather than exhausting its onsite ad inventory.

One former Walmart Media Group executive estimated in a September 2023 interview that Walmart.com has an ad capacity of $3-5 billion, which is only 6% to 9% of FY23 Walmart US e-commerce revenue of $53.4bn. If Walmart wants to continue growing ad revenue to $10 or $15 or $20 billion, it would need to find new avenues, whether in-store or offsite. This suggests a much shorter timeline to potentially flipping retail media ad spend from majority onsite to majority offsite.

Source: eMarketer (November 2023); Bristlemoon Capital

Ad industry expert interviews seem to be in general agreement that retail media will be an important contributor to TTD’s growth in a cookie-less environment once the initial signal loss has been overcome. This is because access to 1^st party retailer data will be much more valuable, especially when combined with advertisers’ own 1^st party data. But before we get too excited about the offsite retail media opportunity, we need to consider what offsite retail media could look like from a ROAS perspective.

In this stylized example, consider an e-commerce website that generates $10 billion of sales and $1 billion in onsite retail media advertising revenue. For the retailer to realize $1 billion of ad revenue, the advertisers would have started with $1.2-1.4 billion of budget depending on total ad tech take rates (20-30% seems reasonable for the agency, DSP, SSP and any data assets that were activated). We assume there was minimal ad fraud or wasted spend given the activity is taking place on the retailer’s own website.

This study by a small Polish martech data science company found that 20% of e-commerce sales were driven by media spend. This appears to include all media spend, not just retail media. 20% does feel low, and if we assume onsite retail media has higher conversion and cleaner attribution due to its proximity to point of sale, we can perhaps generously assume 30% of sales are directly driven by onsite retail media (this is a stylized example after all).

Applying that 30% impact to $10 billion sales implies that $3 billion of sales can be attributed to the $1.2-1.4 billion of gross ad spend, for a ROAS of 2.1-2.5x (and only ~1.5x if 20% of sales were driven by media spend). Now if we leave the onsite environment and move the $1.2-1.4 billion offsite, the ad tech fees increase because there are more middlemen involved; the retailer takes a cut for sharing its data; and ad fraud, brand safety, measurement costs and wasted spend go up. For every $1 the advertiser spends, 50c reaching the end audience as an advertisement might be a good outcome. Offsite click through rate and conversion will almost certainly be lower than onsite due to lack of purchase intent, offset to some extent by lower CPMs. Measurability is also a challenge if the ad viewer takes a circuitous route to buying the product. So that 1.5-2.5x onsite ROAS could quickly become less than 1x in an offsite environment. This is not even factoring in the impact of cookie deprecation on offsite targeting.

Overall, we are not confident that the retail media opportunity will be anywhere near as meaningful as management makes it out to be (i.e., a $500 billion TAM and TTD addressing most of it). However, the more fragmented retail landscape compared to the premium CTV industry does reduce the risk of new walled gardens arising. As the go-to ad buying platform for CTV, TTD is in a strong position to expand retail media into the CTV channel. Also, as TTD will mainly be addressing offsite retail media where brand safety, ad fraud, measurement and data demands are higher, its take rate should also be higher than premium CTV.

Take rate sustainability

The foregoing discussion about TTD’s positioning and opportunity to capture a greater share of digital ad dollars is only one half of the company’s growth equation. The other half is the take rate – that is, the percentage of advertisers’ gross media dollars flowing through its platform that TTD can extract as revenue. This number, unsurprisingly, garners a lot of investor attention, not least because it has remaining remarkably stable in the 19-20% range since the company listed in 2016.

Source: Bristlemoon Capital; Company filings

Company formers we heard from suggested that management is very confident in the capabilities of the platform and in the sales and client success teams' ability to expand client relationships without resorting to discounts. This is apparently quite different to how some other prominent DSPs handle the sales and client service relationship, which range from wining and dining that is typical of Madison Avenue to free credits for use with measurement partners. Instead, pricing decisions are centralized, and sales reps do not have much leeway to negotiate special deals, though will provide volume discounts in line with standard industry practice for the largest clients.

When questioned on the sustainability of the 20% take rate, management offers a generic response along the lines of “it’s always been around 20% and should remain around 20% in future”, and points to all the value unlock the platform has provided over time to support the take rate.

“And we've always communicated that 19% to 20% is what we think is the fair cut for buy side, for the value that we're delivering. And I have, at this point, absolutely no reason to believe that for the foreseeable future, that is going to change…But what we're also doing is, we've had Next Wave, we had Solimar, we have Kokai, and we're building more and more and more and adding more value….We're going to be paid for the value that we're adding. And that, when you combine all of that together with the historical track record, gives me the confidence to say like we're managing to the 19% to 20% and our clients are accepting of that.” – CFO Laura Schenkein, RBC conference November 2023.

The reason why the question of take rate sustainability arises so often is because the ad tech industry as a whole has been notorious for opaque pricing and fee gouging, with each link in the supply chain trying to extract as many dollars as possible from the advertiser’s ad budget. Exacerbating this problem is the fact that any single participant has very little visibility across the entire supply chain. Advertisers and their DSPs only see ad dollars through the bidding stage – once the auction ends and the winning bid amount (less the DSP fees) is passed to the ad exchange or SSP, the demand side loses track of those dollars. The supply side, on the other hand, only sees the net winning CPM, from which the ad exchange, SSP and publisher ad server take their cut. The publisher never sees how much the ad buyer actually bid, and the advertiser doesn’t know how many dollars of its ad budget actually reached the audience in the form of ads.

Source: ANA Programmatic Media Supply Chain Transparency Study: COMPLETE REPORT, 2023

This practice has been exposed by multiple industry studies in recent years and has led to advertisers, agencies and publishers alike applying greater scrutiny to ad tech fees. This has largely been to the detriment of SSPs which have seen their take rates compress faster than DSPs, due to the lesser perceived value of any single demand source to the publisher, as publishers typically have dozens of SSP partners in an attempt to maximize yield. Compare this to the value of a well-connected buying platform to advertisers, who are incentivized to consolidate their ad budgets through fewer DSP partners for volume discounts and more consistent measurement across campaigns. 

Based on our research and conversations, our understanding is that TTD’s ~20% take rate of gross ad spend can be broken down as follows. Note that the company never discloses, even qualitatively, the composition of its eerily stable take rate, leaving investors and even Wall St analysts guessing in the dark as to the makeup of the equal-most important metric for understanding TTD’s earnings power.

• 12-15% base platform fee that is charged to clients for using the basic features of the platform. We believe this is a tiered fee with volume discounts for the largest agency clients. We also believe, despite management’s suggestions otherwise, that the base fee varies by channel, with CTV likely being the lowest on average and digital display being the highest (we explain our reasoning later). One Tegus expert suggested that TTD barely breaks even on the platform fee, and profitability stems largely from incremental data and other value-added fees.

• 5-8% data and value-added feature fees that are charged on an a la carte basis. Features such as TTD’s owned data sets or 3^rd party audience data, campaign optimization, bid shading, other Koa features, clean room access, measurement features etc. all cost incremental fees to turn on.

See Quo Vadis’ excellent post on TTD’s fees that suggest the take rate could be an even higher percentage of an advertiser’s media budget. Consider that TTD’s reported take rate of 20% is already higher than the DSP fees cited in the above industry fee studies (corroborates TTD’s positioning as a premium (i.e., more expensive) DSP), and one can understand why this topic garners so much debate.

We believe there are two main factors that could influence the take rate going forward: 1) the impact of CTV mix, and 2) the deprecation of 3^rd party cookies in Chrome.

Impact of CTV mix

As discussed in the CTV section, DSP take rates can vary substantially depending on the programmatic model being used, with PG as low as 3% and open auctions as high as 20%+. We were unable to source any authoritative opinions on how TTD’s CTV take rate compares to the blended total take rate of ~20%. Triangulating all the data points, we believe TTD’s CTV take rates may be in the mid-teens percentage range. This is for several reasons:

1. Mathematically, a lower CTV take rate most readily explains the historical stability of TTD’s total take rate. Customer concentration has reduced over time, with three >10% customers accounting for 43% of gross spend in 2017 falling to two >10% customers accounting for “>20%” of gross spend in 2022. Direct brand relationships have also expanded during that period; these brands should have less bargaining power/volume discounts than the agencies aggregating the ad spend of multiple brands. TTD has also launched numerous new products, capabilities, and data integrations since 2016, and we know from user feedback that TTD nickel and dimes for just about every single feature outside the basic platform functionality. Putting these together – less customer concentration and more a la carte features – should have driven the take rate higher over time, yet it has stayed flat. At the same time, CTV has gone from basically zero in 2016 to nearly 50% of gross spend in 2023. Management says they are simply providing more value to clients by adding all these extra features and keeping the take rate constant. Occam’s razor would support the argument that CTV carries a lower take rate than other channels.

2. CTV is a cookie-less environment, which means: 1) there is less 3^rd party data being collected in the channel; and 2) there is potentially less 3^rd party data being applied to inform audience targeting due to the unavailability of cookies. Perhaps this means that what data is available become more expensive, but we wouldn’t preclude the possibility that the CTV channel just comes with lower data fees. One Tegus expert mentioned that even if there’s less data fees, there are still fees for brand safety and measurement tools, though we would note that advertisers use these tools for open web display as well, so these fees are not exclusive to CTV.

3. PMP auctions should in theory be lower fee than open auctions because the advertiser and publisher typically have a direct relationship, particularly in concentrated channels such as CTV (relative to the open web). As we discussed in the Retail Media section, publishers and advertisers should really only be willing to pay high ad tech fees if their ad tech partners are bringing incremental demand and supply that they cannot easily or efficiently source themselves. We don’t see why the CTV channel should be any different.

4. While we don’t subscribe to Redburn’s view that substantially all of TTD’s CTV volume is decisioned programmatic, we concede this is the most plausible explanation for the company’s ~20% overall take rate. Every point of CTV gross spend that is PG at 3-5% take rate instead of decisioned programmatic at potentially 15-20% take rate needs 3-4 points of other gross spend at 25%+ take rate to average to a ~20% take rate. If TTD’s CTV gross spend mix was proportional to the industry’s programmatic mix – that is, ~43% PG and 57% decisioned (recall the 56%/19%/25% direct/PG/decisioned chart from the Investor Day) - then 60% to 80% of TTD’s total gross spend would need to be at a 25%+ take rate to offset the drag from PG.

We admit that we could be entirely wrong with our CTV take rate assessment, and it could just be 20% across all channels. TTD has ways of massaging or optimizing its take rate around the 20% mark. Take for example the Koa bid optimization tool, which helps ad buyers minimize the CPM bid on first price auctions by automatically adjusting the CPM and/or impression type to still achieve the campaign goals (substantially all exchanges and ad servers shifted from second price to first price auctions by 2019). Koa incurs a fee to activate, claims to save advertisers ~20% on their winning bids, and has a 20% revenue share of the savings[32]. When an ad buyer that doesn’t use Koa bid optimization wins a $10 CPM auction, TTD collects a $2 fee (assuming the average 20% take rate). When the ad buyer does use Koa and wins with an $8 CPM, TTD collects its 20% fee ($1.60) and takes 20% of the $2 saving ($0.40). While the total fee is still $2, the denominator has been reduced to $8, for a 25% take rate (plus whatever bps fee it costs to activate Koa).

However, regardless of historical take rates, it is future take rates that matter for the stock. Going forward, we do find it difficult to see how TTD can maintain a ~20% take rate on CTV (if it is even that high to begin with). This is mainly based on our view that the largest ad-supported CTV services (Netflix, Disney+ and Amazon) will continue to prefer PG for the fixed CPM and greater control over ad placement, for as long as the premium CTV inventory supply/demand imbalance persists. We also discussed how existing direct relationships between advertiser and publisher in the premium CTV market could put a cap on fees that ad tech intermediaries can extract. Finally, we highlighted the risk of agencies bypassing the DSP to transact directly with the publishers, something they’ve managed for decades in the linear TV market. Hence, we see a mid-teens CTV take rate as a more reasonable base case forecast.

3^rd party cookie deprecation

We explore the topic of Google’s planned deprecation of 3^rd party cookies in Chrome and the potential impact on the ad tech ecosystem in a subsequent section. For now, in the context of take rates, we note that cookie deprecation will reduce the amount of 3^rd party data that is collected and available to advertisers on the open web, which may or may not be recovered by alternative identifier solutions. Less 3^rd party data available for activation means less high margin data fees for TTD, which will undoubtedly impact take rates. Even if the non-Google ad tech ecosystem rallies around alternative IDs and recovers the signals lost to cookie deprecation, it is a process that is likely to take several years.

Because Chrome cookie deprecation only commenced in January 2024 for 1% of Chrome users, it is still too early for us to find any reliable information regarding the impact on DMPs, 3^rd party data activation and take rates for DSPs. This is a development we will follow with interest given its potential impact on TTD, the broader ad tech ecosystem, and the walled gardens.

Pulling everything together, if we assume the average CTV take rate is mid-teens, then the blended take rate for non-CTV channels must be low- to mid-twenties considering 40% to 45% of gross spend is CTV. With the recent entry of the SVOD heavyweights into ad-supported streaming, there will be an influx of premium PG inventory over the next few years. We see downward pressure on TTD’s CTV take rates over this period, or lower gross spend growth if TTD elects not to participate in this PG inventory. We also expect pressure on the non-CTV take rate stemming from Chrome’s deprecation of cookies, mainly through the degradation of 3^rd party data, but also from the loss of any measurement and attribution capabilities tied to 3^rd party cookies.

Source: Visible Alpha; Bristlemoon Capital. Note: in both scenarios, non-CTV take rate declines from 24% to 20% due to 3^rd party cookie deprecation.

OpenPath and Supply Path Optimization

One initiative that both demand- and supply-side ad tech vendors have started to pursue in the face of mounting fee pressure is Supply Path Optimization (SPO). At its simplest, SPO describes ad tech supply chain consolidation by collapsing or bypassing layers of intermediaries to reduce the layers of ad tech fees and increase the share of media spend that ultimately reach the publishers’ pockets. DSPs have engaged in SPO by integrating directly with publishers (bypassing exchanges and SSPs), while SSPs have sought to do the opposite, giving advertisers and agencies direct access to inventory. The impetus for SPO largely stems from the agencies who are under pressure from advertiser clients to find cost savings, efficiencies and greater transparency into ad spend.

TTD’s SPO initiative is called OpenPath, which was announced in early 2022 and initially integrated with 20 large news publishers (unchanged as of June 2023), with plans to expand to 100 publishers by June 2024. Generally speaking, these are large publishers capable of managing their own ad inventory, measurement and yield optimization. By bypassing the supply-side intermediaries, TTD promotes OpenPath as delivering a better outcome to struggling publishers. Of course, cutting out the ad exchange and SSP fees alleviates the pressure on TTD’s own fees, as total ad tech fees have been reduced. We would not be surprised if TTD also charges clients extra to buy directly from the publisher, reasoning that it eliminates ad fraud, reduces duplication and provides greater transparency into the ad dollar. Because this is such a nascent initiative, TTD has not offered any tangible commentary around OpenPath, though we think they may disclose more through 2024.

Source: The Trade Desk 2022 Investor Day

TTD is by no means the only beneficiary of SPO. Magnite, the largest listed independent SSP, announced ClearLine in May 2023, which gives agencies direct access to premium CTV inventory on the Magnite platform through direct and PG deals. In 2022, agency holdco WPP’s GroupM (the world's largest media buying agency) created Premium Marketplaces by partnering with Magnite and Pubmatic, another listed SSP, to directly access premium inventory across display and CTV without a DSP. Magnite and Pubmatic have established SPO partnerships with other agencies as well. All this is to say that there is no clear winner from SPO. TTD may benefit on one hand from bypassing SSPs, but on the other hand may come under pressure from agencies threatening to integrate directly with inventory suppliers. Given TTD’s scale in the industry, we suspect SPO has a neutral-to-positive impact, though it is hard to say for certain.

The company seems acutely aware of the take rate mix headwinds that are likely to intensify going forward. In addition to OpenPath, management has flagged several areas of future investment that, if successful, could help mitigate the take rate pressures.

1. The Forward Market – a data driven guaranteed marketplace for buying CTV inventory. Buyers can pay a premium to have impressions in the future guaranteed and sellers can offer a discount for guaranteed future revenue, while still maintaining the decisioning and control of an auction.

2. Investing in education to help advertisers understand the relationship between linear and CTV. Most advertisers are very interested in incremental reach, so a key stepping stone to CTV adoption is demonstrating to advertisers how they can extend a linear campaign into biddable CTV to reach incremental viewers.

3. A more flexible data marketplace. Historically, pricing of data didn't have strong correlation with outcomes, so advertisers used (i.e., paid for) fewer data elements in decisioning than TTD would like. Data partners were also incentivized to optimize for scale rather than quality, because quality data wasn’t necessarily reflected in the price or uptake. The data marketplace offers hybrid pricing to better align pricing with results and splits payouts between the data providers based on the incremental value of each data set to campaign targeting. Encouraging greater activation of data is beneficial to TTD’s take rate, though we think there is a limit to how much data can be used effectively in a single ad campaign.

Unified ID 2.0 and the deprecation of cookies

Finally, we cannot conclude a deep dive on an ad tech business without addressing Google’s deprecation of 3^rd party cookies in the Chrome browser. The ad tech industry’s response to the planned deprecation of Chrome cookies has been to develop dozens of alternative deterministic (using emails and device IDs) and probabilistic (based on signals like IP addresses, browser, location etc) identifier solutions. Unified ID 2.0 (UID2) is TTD’s identifier solution for the open web and the most prominent alongside LiveRamp’s RampID. UID2 is a deterministic solution that takes a user email and hashes it into an alphanumeric identifier that can then be tokenized and passed along the bidstream to allow ad tech intermediaries to identify the user pseudonymously.

We will spare the details of how UID2 works and only highlight what is important for investors to know. TTD wants UID2 to become an industry standard ID solution and insists it is not a TTD solution even though the company designed and administrates it unilaterally. TTD had initially roped independent industry bodies Prebid.org and IAB into operating and administrating UID2, respectively, but both pulled out over what could ostensibly be described as independence and privacy concerns. This has not deterred TTD from pushing ahead with the broad rollout of UID2, which has expanded to over 500 partners spanning agencies, publishers, ad tech vendors and advertisers.

“Levels of activation that high [75% of 3^rd party data ecosystem] mean we will have effectively solved the identity matching challenge of the entire open internet on a scale well beyond anything cookies have ever accomplished, and all while providing consumers with much greater control over their privacy.” – Jeff Green, Q4 2022 earnings call.

In a nutshell, Jeff Green claims to have solved the cookie matching problem – where 3^rd party cookies typically have match rates between 40-60% – and saved the open internet from the deprecation of 3^rd party cookies in Chrome. We think this proclamation is self-serving and premature. Self-serving because UID2 only works for publishers that have a logged-in audience, and we noted earlier that as low as 5% of open internet website users are logged in. TTD claims to fight the good fight for the open internet and UID2 makes great PR spin, even though it doesn’t help much of the open internet. Premature because substantially all ad tech expert interviews we heard have expressed a view that UID2 will be just one part of an entire ecosystem of 3^rd party alternative identifiers, with no single solution capable of replacing what is lost to cookie deprecation. UID2 garners a lot of industry attention not necessarily because it is the best solution, but rather because TTD has put its marketing muscle behind it.

We also heard from multiple sources that UID2 adoption rates and matching rates are not as high as TTD had hoped and thus can’t handle high volume campaigns. We see this potentially as more of a timing issue. Google finally started deprecating cookies in January 2024, starting with 1% of users with full deprecation by year-end, after three years of delays. These delays led many in the open web advertising ecosystem to ignore or push back solving the challenge of cookie-less targeting. Now that cookie deprecation has finally started, we suspect there will be a mad rush to adopt alternative identifiers, of which UID2 is a prominent one. Instead, we see two main challenges with UID2: privacy and signal.

Privacy issues

When a publisher decides to implement UID2, logged in users should be presented with a consent notice that explains the “value exchange” of the open internet, i.e. you let us use your email address for targeted advertising and we let you consume our content (sometimes for free). If the user consents, their email address is turned into a UID2 identifier that is maintained by the UID2 administrator (which remains TTD since it can’t find an independent administrator). We say “should” because we are yet to see a notice explaining this value exchange despite signing up to several publishers that are listed as UID2 partners (we were only presented with cookie consent notices). Presumably, publishers are hiding the notice deep in their website T&Cs or privacy policies, since the CCPA (as amended by the CPRA) doesn’t require explicit consent, unlike Europe’s GDPR.

Source: Unified ID 2.0 developer documentation

In our opinion, UID2 is one of the most privacy invasive alternatives to 3^rd party cookies that anyone could conceive. As a deterministic ID, a user’s UID2 is unique and tied to their email address. This means their UID2 follows them around everywhere. Cookies are tied to a digital activity in a browser, and even Apple’s IDFA is tied to a device; if a user changes browsers or devices, the identity link is broken. Not so with UID2, as any activity where a user enters their email address could be conceivably linked back to their UID2.

Say you pick up some items from Target with your loyalty card on the way home from work in an Uber, then crash on the couch to watch some Disney+ while ordering dinner on DoorDash, then scroll some Buzzfeed articles and play five minutes of the new AppLovin game as you wait for your food, then decide you’ve been a slob and should spend 30 minutes on your connected exercise bike, and finally jump on your laptop to post some comments on the [insert activity] forum you frequent. If each of these entities have implemented UID2, you’ve used the same email to register for all of them, and you’ve given implied consent by accepting the privacy policies you didn’t read (since CCPA doesn’t require explicit consent for adults), then the data from each activity you engaged in can be linked specifically to you as an individual, regardless of whether it was in the physical or digital world, on your TV or your phone or another connected device.

We must ask – in what world is this more privacy safe than 3^rd party cookies that are contained within a browser on a single device? The privacy spotlight is on big tech right now, but as and when regulators turn their attention to alternative identifiers, it is difficult to see how deterministic identifiers such as UID2 can pass regulatory scrutiny. Google is imploring the open web ecosystem to test its Privacy Sandbox before jumping on the alternative ID bandwagon, because it believes alt IDs won’t meet ever-stricter privacy regulation. Of course, Google is incentivized to say that, but that doesn’t mean it is wrong.

“I would be skeptical and really reassessing companies who are doubling down on a different identifier to cookies, that could be viewed as even more invasive from a privacy perspective. Hashed e-mail, IP addresses, things like that. I think because cookies have become under the scrutiny that they have, those [alternatives] are much more deterministic, much closer tied to PII and identifying an ID back to a real human, so I am certainly not a fan of doubling down on that deterministic identity strategy. I think the people who are investing there are eventually in for a rude awakening.” – Former SVP at MediaMath Inc.

Signal loss

Even if Target, Uber, DoorDash, Disney+, Buzzfeed, AppLovin, your favorite special interest website or any of the 500+ partners have implemented UID2, the solution is useless without 1) user opt-in; and 2) some way of sharing the data tied to an UID2. If users don’t opt-in to their emails being turned into UID2 identifiers, then it doesn’t matter that a publisher with millions of logged-in users has implemented UID2. Statista shows 3^rd party cookie consent rates being in the 32% to 64% range by country[33], while Adjust’s data for Apple’s App Tracking Transparency (ATT) show less than 35% consent rates industry wide[34].

As we mentioned earlier, we’ve never come across a prompt asking us to consent to the “value exchange” of the open internet, so it could be buried in website/app privacy policies. Interestingly, the CCPA requires explicit consent for 3^rd party cookies, but not for an ID solution that turns your email into a unique identifier that follows you around everywhere you go. Obviously, if no explicit consent is required, “opt-in” rates for UID2 will be very high – which likely explains why Jeff Green is so confident in UID2 adoption. If consent is required, we don’t see why the opt-in rate would be meaningfully different to existing 3^rd party cookie or ATT opt-in rates.

Once consent is obtained, UID2s are created and the data is flowing, that data still needs to be shared throughout the ad tech ecosystem to facilitate targeting. A UID2 is just an identifier that doesn’t contain any information; passing UID2s around doesn’t pass along the underlying user data. The difference with 3^rd party cookies is that a 3^rd party (typically a DMP or some other ad tech intermediary) is placing a cookie and collecting data on a publisher’s website, and it is in their economic interest to be selling the data. With UID2, the publisher is collecting 1^st party data and needs to share it with the advertiser or ad tech intermediaries to make it available for use in audience targeting. If publishers and advertisers don’t share the data, then everyone is stuck with their own 1^st party data and left waiting to match UID2 impressions with existing UID2s they have on file.

Interestingly, Jeff Green made the following comment on the Q3 2023 call: “With UID2, content owners don't have to share data. Instead, advertisers can use their own data, and to do so, they're willing to pay more.” This is of course referring to the narrow objective of retargeting, whereby advertisers target audiences that have previously engaged with them in some way. John is a Nike member, and a Nike ad campaign receives a bid request with a UID2 that matches John’s. Nike now knows that John is viewing the ad but doesn’t know any more about John than what it already had on file. We struggle to see how an advertiser could use UID2 to run a campaign targeting new audiences without data sharing across the advertising ecosystem. It also does not appear to be in TTD’s interest to promote less 3^rd party data usage considering it earns very high margin data activation fees.

Because of the Swiss cheese-like coverage of UID2 – recall the statistic that only 5% of the open web is logged in – we believe that some, if not most, open web audience targeting is still informed by data from 3^rd party cookies. UID2 provides an extremely porous view of user activity, tracking only activity at points where users are logged in. Cookies, notwithstanding the syncing challenges that lead to just 40% to 60% match rates[35], can track users as they navigate across the open web.

Impact of cookie deprecation

On the Q2 2023 earnings call, Jeff Green has fielded a number of questions on the impending deprecation of cookies in Chrome and expressed a view that the impact on TTD is likely to be neutral to slightly positive.

“A lot of the speculation reminds me of the frenzy in 2017 when Safari deprecated cookies, or more recently, in 2021, when Apple made its IDFA changes. Each time I have stated that none of this will have a significant material effect on our business, and that was true.”

“So I do think that it would accelerate the adoption of UID2 into the browsing internet world, not just among CTV and audio, and that would ultimately be good for us…so it would be a net negative for the open internet, but probably a slight positive for us.”

“So we look at 12 million ad opportunities every second, and we choose from those 1 million or 2 million that we want to buy. If identity is removed from instead of it being on 6 million, it's now on 3 million, it just informs which ones we buy differently.”

We accept that cookie deprecation likely has minimal impact on CTV advertising because CTV is a cookie-less environment to begin with (though we wonder if fingerprinting e.g. IP addresses are used to match CTV viewers to web browsing activity in the same household…). Where we don’t share Mr Green’s confidence is in the belief that cookie deprecation will have no impact on TTD’s open web ad buying business. Even though CTV is approaching 50% of gross spend, the remaining mix is predominantly mobile web and display advertising which are heavily dependent on 3^rd party cookies and likely have limited UID2 adoption compared to CTV.

Sure, TTD only bids on a fraction of the 12 million QPS it sees and can shift the mix of impressions it buys based on identifiability of the audience. But if the platform is already bidding on the most optimal 1 or 2 million impressions available based on cookies and alternative IDs (as it should be if its optimization algos are working), the deprecation of cookies should in theory reduce the overall signal and value of all impressions as a whole. Therefore, regardless of how TTD shifts its buying, CPMs on the open web as a whole should decline. We already discussed in the take rate section why we believe TTD’s open web take rate may be higher than CTV take rate, so this decline in CPMs and thus gross spend in the higher take rate channels can have a meaningful impact on TTD’s financial performance.

When we searched for empirical studies of the impact of cookie elimination on CPMs, we found only one comprehensive academic study[36] comparing the difference in CPMs between Chrome and Safari browsers between 2017-2020, during which Apple gradually phased out 3^rd party cookies and ad tracking entirely from the Safari browser under Intelligent Tracking Prevention (ITP).

The study uses a data set of up to 2 billion programmatic auctions per day across 1,100 advertisers, over seven 90-day periods around each ITP version release. To our surprise (and the authors’), the study found that Safari CPMs were only -5% lower than Chrome because of ITP, a far cry from industry research showing Safari CPMs being 50% to 60% lower than Chrome post-ITP. What is interesting though is the heterogeneity in the data – while the modal impact of ITP on open auction (RTB) CPMs was clustered around zero, the mean and median impact appears to be negative, at least for programmatic display ads. The study also reviewed past empirical literature on the impact of cookies on programmatic CPMs, which found that CPMs for cookie-less impressions were up to 50% to 70% lower than for impressions with cookies. The findings in the study give us sufficient confidence that open web CPMs will most likely be lower post Chrome cookie deprecation.

Source: Duckworth et al (2023)

Even intuitively, in the context of UID2, we pose the following question: given Safari has been eliminating 3^rd party cookies since 2017, and UID2 and other alternative IDs have existed for several years, why are we not observing Safari CPMs to be massively higher than Chrome CPMs at publishers who have implemented UID2? TTD’s spin is that cookies are unreliable and UID2 solves the identity matching problem for the open internet. Safari has no cookies, but definitely has millions of users with email addresses that can be turned into UID2s. And conventional wisdom is that iOS/Apple users are more valuable than Android/Chrome users (judging by Apple and Google’s respective App/Play Store revenues), so advertisers should be willing to pay more to reach Safari users even without the benefits that UID2 claims to bring.

(As an aside, Apple’s deprecation of the IDFA had a negligible impact on TTD because only 10% of QPS had an IDFA attached, per a comment on the Q3 2020 earnings call. We suspect the 10% IDFA impressions was much lower than the 40%+ mobile share of gross spend at the time because TTD isn’t a big buyer of app-based ad impressions and skews more heavily to mobile web where there are less walled gardens. Where IDFAs were no longer available, TTD redirected bids to the other millions of mobile QPS that weren’t affected. This won’t be feasible for Chrome cookie deprecation because it will affect substantially all mobile QPS that still have cookies attached. This means the privacy headwinds for TTD are still ahead of it, while for a company like Meta, the privacy challenges are largely behind them.)

Valuation considerations

In a break from our previous deep dives, we are not providing a model or valuation for TTD as there is simply too wide a scope of possible outcomes to keep the discussion brief. This deep dive has already reached bathysphere levels of depth (congratulations on making it this far) and throwing in our forecasts at the end is not additive to the reader’s understanding of TTD as a business and the opportunities and challenges that lie ahead.

Instead, we provide our high-level thoughts on some key factors that readers should consider when forming their own forecasts and views on valuation.

• CTV growth rates – given TTD’s already high share of decisioned programmatic CTV and the influx of premium PG inventory from the largest streaming services (several within walled gardens), we believe TTD may see a deceleration of growth relative to the broader CTV industry, especially if management elects not to pursue low-fee PG spend. If TTD does increase its PG mix to maintain growth, the 20% take rate will come under pressure. We are less concerned about CTV growth rates as we are about the 20% take rate.

• Take rate development – as we laid out in the take rate section, we believe CTV is lower take rate than the company average and may come under increasing pressure as more linear budgets shift to CTV. Additionally, we don’t see how Google’s deprecation of cookies doesn’t have a disruptive impact on TTD, at least near term, in the form of less data available for audience targeting.

• Margin development – the key thing readers need to be aware of on the margin front is that “Platform operations” (essentially COGS) is a variable expense tied to the number of impressions (QPS) and bids the platform handles, not to revenue. This means that as the take rate declines, we can expect to see gross margin deleverage. Additionally, sales and marketing expense has seen deleverage on both a GAAP and non-GAAP basis over the last four years despite revenue more than tripling. This may be another source of margin pressure as TTD ramps brand direct relationships, international expansion, and marketing efforts to educate advertisers on CTV, UID2 and OpenPath.

• Share based compensation – SBC has seen a massive increase since 2021, jumping to over 30% of revenue from low-teens in 2019-20. This is largely due to a massive CEO Performance Option that was awarded to Jeff Green in 2021, with a grant date fair value of $819 million. The options vest in eight equal tranches over 10 years at stock price triggers of $90 to $340, though the expense profile is front-loaded, with $158 million and $262 million expensed in 2021 and 2022, respectively, and a further $399 million to be substantially expensed over 2023 and 2024. Accordingly, non-GAAP operating margin including SBC has shrunk to just 8.5% for the trailing 12 months to Q3 2023, from 22.5% in 2018. We typically treat SBC as a cash expense, though the (hopefully) non-recurring nature and accelerated expensing of the CEO Performance Option can arguably be excluded from earnings. We also take a dim view of massive option packages for CEOs to align their incentives with shareholders, especially for a billionaire CEO who already owns nearly 10% of the company and has 48% voting power.

Source: Bristlemoon Capital; Company filings

• International expansion opportunity – we have not devoted any text in this report to TTD’s international expansion opportunity because 1) it does not appear to be a priority for management; and 2) there has been no sustained progress for the past 10 years, with international gross spend mix bouncing between 10% and 15% of total and declining in recent years due to rapid growth of CTV in the US. One advantage over Google and Meta that management likes to highlight is the fact that TTD can operate in China. However, we note TTD first entered China in 2016 and officially launched its DSP in 2019, yet there hasn’t been any progress on the international mix since (granted China was severely disrupted by COVID followed by geopolitical issues). Accordingly, international expansion is not something we would explicitly pay upfront for, until there is more evidence of sustained progress being made.

Source: Bristlemoon Capital; Company filings

• Google ad tech antitrust lawsuit outcomes – finally, it would be remiss of us not to mention the pending Google ad tech lawsuit. Should Google be ordered to separate its ad serving technology from its ad exchange to prevent it self-preferencing Google properties in an auction, it would greatly level the playing field for independent ad tech. Several expert interviews suggest that 20% to 50% of Google’s Network ad revenue ($31 billion in 2023) could be up for grabs, with TTD being the biggest beneficiary. Before anyone gets too excited by the prospect of $6 billion to $15 billion of ad revenue being shared across independent ad tech, we would remind that Google recognizes Network revenue as principal (i.e. gross), so that $31 billion of “revenue” is actually gross spend from the perspective of TTD. This is still a huge amount of market share potentially up for grabs, just less extreme than at first blush. On the downside, if Google decides to spin off its Network business altogether (considering revenue has declined for six consecutive quarters), DV360 could become a more formidable competitor to TTD in CTV ad buying. We understand that part of the reason why Google has underinvested in its CTV ad tech stack is because of the very real conflict of interest with also owning YouTube, the largest supply of CTV inventory in the US. Should the ownership link between DV360 and YouTube be severed, more advertisers could get comfortable using DV360 for their CTV campaigns.

Source: Bloomberg consensus; Bristlemoon Capital

Ultimately, TTD’s valuation looks stretched to us compared to other businesses in the ad tech industry and even other high growth software businesses. The Trade Desk platform is by most accounts a fantastic product in terms of UX, capabilities, transparency and control. The TTD business has several attractive growth opportunities even though management is perhaps overly promotional about the company’s prospects. The TTD stock, however, appears to be priced to perfection when in our view there is a level of uncertainty and risk associated with the growth opportunities that investors are not being compensated for.

Disclaimer / Disclosures

The information contained in this article is not investment advice and is intended only for wholesale investors. All posts by Bristlemoon Capital are for informational purposes only. This article has been prepared without taking into account your particular circumstances, nor your investment objectives and needs. This article does not constitute personal investment advice and you should not rely on it as such. This document does not contain all of the information that may be required to evaluate an investment in any of the securities featured in the document. We recommend that you obtain independent financial advice before you make investment decisions.

Forward-looking statements are based on current information available to the author, expectations, estimates, projections and assumptions as to future matters. Forward-looking statements are subject to risks, uncertainties and other known and unknown factors and variables, which may affect the accuracy of any forward-looking statement. No guarantee is made in relation to future performance, results or other events.

We make no representation and give no warranties regarding the accuracy, reliability, completeness or suitability of the information contained in this document. To the maximum extent permitted by law, we do not have any liability for any loss or damage suffered or incurred by any person in connection with this document.

Bristlemoon Capital Pty Ltd (ABN: 22 668 652 926) is an Australian Financial Services Licensee (AFSL Number: 552045). 

George Hadjia is associated with Bristlemoon Capital Pty Ltd. Bristlemoon Capital may invest in securities featured in this newsletter from time to time.

---

[1] https://abtc.ng/jeff-green-net-worth-how-much-is-the-billionaire-who-has-quit-the-mormon-church-worth/

[2] https://www.forbes.com/sites/amitchowdhry/2017/12/12/how-jeff-green-took-the-trade-desk-from-a-simple-idea-to-a-programmatic-ad-giant/?sh=27f712a811da

[3] Ibid.

[4] Ibid.

[5] https://www.adexchanger.com/investment/ad-exchange-adecn-acquired-by-microsoft/#:~:text=Feeling%20it%20couldn't%20be,the%20%2450%2D75%20million%20range.

[6] Ibid, 2.

[7] Ibid, 2.

[8] https://www.fa-mag.com/news/ad-man-inspired-by-goldman-becomes-billionaire-with-trade-desk-40613.html?print

[9] https://tracxn.com/d/companies/the-trade-desk/__NO68rlFZ3HkX9PHbfLVRBsnFK0tWFRNqcjMtJWp_PxY/funding-and-investors

[10] https://www.adexchanger.com/advertiser/the-trade-desk/

[11] https://www.thetradedesk.com/us/news/press-room/the-trade-desk-launches-td7-venture-capital-arm-to-invest-in-open-internet-start-ups

[12] Jounce Media, The State of the Open Internet 2023

[13] https://digiday.com/marketing/sources-mediamath-is-exploring-a-sale-magnite-and-pe-are-among-potential-acquirers/

[14] https://www.marketecture.tv/programs/index-exchange-andrew-casale

[15] https://adage.com/article/marketing-news-strategy/how-procter-gambles-media-strategy-changes-role-agencies/2473956

[16] https://variety.com/vip-special-reports/the-fast-approach-to-streaming-content-a-special-report-1235302920/

[17] https://www.pewresearch.org/short-reads/2021/03/17/cable-and-satellite-tv-use-has-dropped-dramatically-in-the-u-s-since-2015/

[18] https://www.hollywoodreporter.com/business/business-news/netflix-disney-now-pushing-subscribers-to-ad-tiers-1235572459/

[19] https://www.insiderintelligence.com/content/what-narrowing-range-of-streaming-service-cpms-mean-advertisers

[20] Also corroborated by Salon, Snopes and TV Tropes: https://digiday.com/media/some-publishers-are-starting-to-see-revenue-lift-from-alternative-ids/

[21] https://www.freewheel.com/wp-content/uploads/2022/10/FreeWheel_VMR_H2_2020_Digital_FINAL.pdf

[22] https://www.freewheel.com/wp-content/uploads/2022/09/FreeWheel_VMR_1H_2022_Digital.pdf

[23] https://digiday.com/media/netflix-is-reviewing-its-ad-strategy-considering-build-or-buy-pivots-away-from-microsoft/

[24] https://www.adexchanger.com/digital-tv/amazons-deals-with-the-trade-desk-and-dataxu-bring-rtb-to-ctv/

[25] https://www.insiderintelligence.com/chart/262152/us-tv-connected-tv-ctv-ad-spending-2021-2027-billions

[26] https://www.freewheel.com/wp-content/uploads/2023/10/FreeWheel_VMR_1H_2023_Digital.pdf

[27] https://digiday.com/future-of-tv/its-not-entirely-clear-what-direct-even-means-ctvs-rise-is-not-without-its-growing-pains/

[28] https://www.adexchanger.com/digital-tv/disney-ups-the-ante-on-targeting-and-measurement-ahead-of-the-2023-upfronts/

[29] https://setupad.com/blog/types-of-programmatic-deals/

[30] https://www.insiderintelligence.com/content/new-data-linear-tv-fell-below-50-viewing-share-july-first-time

[31]

[32] https://www.adexchanger.com/online-advertising/how-the-trade-desk-has-evolved-for-the-next-stage-of-dsp-growth/

[33] https://www.statista.com/statistics/1273012/consent-cookies-worldwide/

[34] https://www.adjust.com/blog/app-tracking-transparency-opt-in-rates/

[35] https://www.publift.com/blog/what-is-cookie-syncing

[36] “Taking the biscuit: how Safari privacy policies affect online advertising”, Simeon Duckworth, Mateusz Mysliwski, Lars Nesheim. January 2023.