The Trade Desk — Where Should the Next Capacity Go?
Download case PDF (7 pages)Request teacher’s pack
Decision date and information cutoff: February 15, 2024, 12:10 p.m. ET
Decision scope: Hypothetical FY2024 allocation of incremental engineering and partner-enablement capacity
Case convention: This is an author-created decision exercise. The role, capacity constraint, alternatives and any proposed targets are hypothetical. Public company statements are identified as such; no private deliberations are represented.
The decision
At 12:10 p.m. Eastern on February 15, The Trade Desk (TTD) had filed its FY2023 Form 10-K just minutes earlier. You advise a hypothetical product leadership team deciding where the next increment of fixed FY2024 engineering and partner-enablement capacity should go. The company has not disclosed a fixed budget or a choice among the initiatives below. Treat the capacity constraint as a teaching assumption, not a reported plan.
Choose a primary emphasis among (A) Kokai, the AI-assisted decisioning, measurement and workflow platform; (B) UID2/EUID, TTD's alternative-identity frameworks; (C) OpenPath, its direct publisher-connection initiative; or (D) a stage-gated portfolio. You may recommend a primary bet with limited options, but specify what gets the first teams, what is deferred, and what evidence would change the allocation.
Use only information available by 12:10 p.m. ET on February 15, 2024. The 10-K was accepted by the SEC at 12:04:49 p.m. ET and is included. TTD's same-day Q4 earnings release, prepared remarks, conference call, and all later materials are excluded. [S1]
Your recommendation should identify a customer or ecosystem outcome, explain how TTD could capture value, account for the stakeholders who may bear costs, and propose a credible test. Aggregate platform growth is not evidence of any initiative's return.
A demand-side platform in a multi-sided market
TTD provides a self-service demand-side platform (DSP) that agencies, advertisers and service providers use to plan, bid for, optimize and measure campaigns across digital channels. In TTD's description, clients choose inventory and control campaign spending; TTD generally earns a platform fee tied to purchases through the platform and may charge for data and other services. It does not buy media to resell it for its own account. The company competes for advertiser use while depending on publishers, supply-side platforms (SSPs), exchanges, identity providers, browsers and other partners. This simplified map omits some routes and intermediaries. [S1]
Digital advertising's supply chain has changed as exchanges and publishers introduced header bidding and other ways to run auctions. More direct or transparent routes may improve information and competition, but can also alter fees, latency, access and bargaining power. In January 2023, the U.S. Department of Justice alleged that Google used acquisitions and auction design to entrench its ad-tech products. Those were pending allegations at the cutoff, not adjudicated findings and not evidence about TTD's conduct. [S5, S6]
Growth without product-level returns
TTD reported 2023 revenue of $1.946 billion, up 23%; gross spend of $9.611 billion, up 24%; and net income of $178.9 million, compared with $53.4 million in 2022. TTD defines gross spend as client purchases plus its platform fee, and notes that other companies may define the term differently. Revenue divided by gross spend is about 20.25% in 2023 versus 20.38% in 2022, an author-calculated aggregate proxy, not a contractual rate, product margin or stable unit price. Client and product mix may affect it. [S1]
| Selected measure | 2022 | 2023 | What it does and does not show |
|---|---|---|---|
| Revenue | $1,577.8m | $1,946.1m | Reported; +23% |
| Gross spend | $7,741m | $9,611m | Company-defined; +24%, not TTD revenue |
| Net income | $53.4m | $178.9m | Reported; not a product return or recurring operating measure |
| Technology and development expense | $319.9m | $411.8m | Reported; +28.7%, with no product allocation |
| Platform operations expense | $281.1m | $365.6m | Reported; +30.1%; includes hosting, traffic/query volume, data, support, personnel and amortization |
Source: TTD FY2023 Form 10-K, Item 7 and financial statements. Expenses are rounded from reported figures. Qualifying software development costs may be capitalized and later amortized in platform operations. Neither expense line identifies investment in Kokai, UID2/EUID, OpenPath or maintenance work. Hosting expense increased by $54 million during 2023, according to management. These aggregates do not identify whether higher costs produced better service or returns. [S1]
TTD's client agreements generally renew annually and can be terminated on 60 days' notice; they do not require material purchases. The company also disclosed that Publicis would exceed 10% of gross billings if its underlying agencies were aggregated at holding-company level in 2023 and 2022. This is concentrated exposure across agencies, not one direct contracting entity. Supplier relationships also generally lack long-term commitments. Adoption, service quality and access therefore have to be sustained on both sides of the platform. [S1]
Three connected product propositions
Kokai: make decisions and workflows better. In June 2023, TTD launched Kokai, describing distributed AI, measurement features and a new interface intended to help clients plan and optimize media. Its launch release said an alpha Partner Portal had helped more than 400 partners integrate over two quarters. These are management descriptions and an integration count, not evidence of active advertiser adoption, incremental campaign lift or durable client value. [S2]
For TTD, the potential chain is improved prediction or workflow → more effective campaigns and easier use → greater client use or spend → net revenue and retention, less added compute, support and engineering cost. A key question is whether the system improves incremental outcomes or mainly predicts who would have purchased anyway. Advertisers may value attributed conversions, but attribution alone cannot answer that question.
UID2/EUID: preserve addressability with different identifiers. TTD describes UID2 as an open-source framework that transforms email addresses or phone numbers into pseudonymous identifiers subject to controls; EUID entered limited beta in Europe in 2023. The strategic thesis is that advertisers and publishers can preserve some relevant reach and measurement as browser identifiers change. But TTD does not report active reach, match rates, partner concentration, opt-outs, consent quality or product-level economics in the available materials. A framework's availability does not establish broad adoption, valid consent or consumer benefit. The company said browsers and platforms control identifiers and described Chrome's planned cookie changes as uncertain at this cutoff. [S1]
OpenPath: change the route to publisher inventory. TTD describes OpenPath as a way to simplify direct connections between its DSP and publishers. The thesis is that a more direct path may reduce friction, improve information or make inventory more accessible. Yet a new route can redistribute economics and influence auction competition. Advertisers may lose breadth or see duplicated paths; publishers may gain access or lose competing demand; SSPs may be bypassed or become less useful. TTD disclosed no OpenPath-specific price, route quality, publisher yield, latency or advertiser outcome. [S1]
The initiatives may complement each other. Kokai could depend on good inventory and usable signals; identity can support decisioning; direct connections can change the data and auction environment. That complementarity also creates dependencies outside TTD's control and makes it harder to attribute results to one investment.
What can the evidence establish?
Research on advertising cautions against equating attributed outcomes with incremental effects. Frick, Belo and Telang studied 208,538 consumers at one European online electronics retailer over 49 days, randomly assigning retargeting or public-service ads. The randomization identifies the ad effect; the further inference that observed bidding favored consumers with higher baseline purchase propensity over responsiveness depends on bid/model analysis, an assumption about platform bidding and imprecise interaction estimates. Their modeled contract comparison is not an experimental TTD return. [S7]
Lewis and Rao's abstract describes 25 large field experiments and the difficulty of measuring return even with very large samples. This source was available only at abstract level for this case dossier. The practical implication is to define the outcome, horizon and statistical uncertainty before a pilot begins—not to accept an attribution metric as proof. [S8]
On supply-path information, Ada, Abou Nabout and Feit studied a private European exchange's staged disclosure of ad-subdomain context. Across 57 sites observed in both years (3,058 site-week-year observations), their controlled difference-in-differences estimate associated full disclosure with €0.111 more exchange revenue per thousand impressions (standard error €0.043). They report identification assumptions and imprecise site-specific results. That study suggests a possible information mechanism under one exchange design; it does not estimate OpenPath's effect on advertiser value, publisher yield or TTD's share. [S9]
The cited evidence gives reasons to test incrementality, information and ecosystem effects. It does not provide a TTD-specific causal lift, identity adoption curve, route-level economics or product return. TTD's 23–24% aggregate growth cannot fill those gaps.
Your alternatives
A. Prioritize Kokai. Concentrate capacity on decisioning, measurement and workflow. This may deepen the value proposition for existing clients and support spend across channels. The countercase: better prediction may not mean more incremental value; implementation, compute and support costs may rise; clients may not adopt or may not share the economics.
B. Prioritize UID2/EUID. Concentrate on partner onboarding, interoperability, controls and coverage. This may preserve addressability and measurement as identifiers change. The countercase: adoption, consent and browser/platform choices are external; multiple identity and contextual approaches compete; an open framework may create ecosystem value that TTD cannot capture.
C. Prioritize OpenPath. Concentrate on integrations and route quality. This may simplify access and improve the match between advertiser demand and publisher inventory. The countercase: directness alone does not prove efficiency; routing can change competition, fees, latency, inventory breadth and publisher net proceeds.
D. Stage-gate a portfolio. Fund small, separately measured work in more than one initiative, with named owners and additional capacity released only against evidence. This can preserve options and acknowledge complementarity. It can also fragment scarce specialists, slow execution and make accountability difficult. A portfolio must state what is not funded and how shared dependencies are handled.
These are hypothetical resource-allocation choices, not disclosed TTD plans. The public record does not identify an optimal allocation or project-level hurdle rate.
Assignment
Prepare a concise recommendation to the hypothetical product leadership team. State:
- Which initiative or customer outcome gets priority, and what work is explicitly deferred?
- How could the investment create value for advertisers, publishers or consumers—and how could TTD capture enough of that value to justify its costs?
- What 2–3 measures and comparison design would test the thesis? Address adoption and quality, not just integrations or attributed conversions.
- What are your scale, redirect and stop conditions? Who bears the cost if your thesis is wrong?
- What is the most important unresolved assumption, and how would your recommendation change if it fails?
Any staffing envelope, dollar budget, forecast, performance threshold or internal plan you introduce is a teaching assumption. Make its unit, timing and rationale explicit. Do not present a company budget, customer example or private meeting as fact.
Discussion questions
- Which initiative addresses the most important bottleneck, and what dependencies must work for it to create value?
- How should TTD distinguish product adoption, attribution and incremental advertiser value?
- How could a direct supply path redistribute value among advertisers, publishers, SSPs and TTD? What would a fair test observe on each side?
- What does the cookie roadmap imply for UID2/EUID, and what remains outside TTD's control?
- When does a gated portfolio preserve option value, and when does it simply dilute accountability?
Sources and conventions
S1. The Trade Desk, FY2023 Form 10-K, accession 0001671933-24-000014, accepted Feb. 15, 2024, 12:04:49 p.m. ET. Business model and clients: printed pp. 4–8; risks, privacy, identity and cookies: pp. 14–39; MD&A, gross-spend definition, product descriptions, expenses and trends: pp. 44–54; financial statements and notes: pp. 60–80. See also filing index and acceptance metadata.
S2. The Trade Desk, “The Trade Desk Launches Kokai”, June 6, 2023. Product design, Partner Portal integration count and staged launch discussion. Company claims are not independent outcome measures.
S3. PubMatic, Q3 2023 financial results, Nov. 8, 2023, financial and business highlights. Company-reported measures with PubMatic definitions; not directly comparable to TTD.
S4. Magnite, Q3 2023 financial results, Nov. 8, 2023, summary, channel table and contribution ex-TAC definition. Company-reported seller-side results; not directly comparable to TTD.
S5. OECD, Competition in Digital Advertising Markets, 2020, Chapter 2, printed/PDF pp. 18–26. Secondary synthesis on auction sequence, header bidding, latency and supply-chain structure; not a TTD impact study.
S6. U.S. Department of Justice, complaint in United States v. Google LLC, No. 1:23-cv-00108 (E.D. Va.), filed Jan. 24, 2023, pp. 1–8, 68–74/Fig. 13. Allegations were pending at the cutoff. See also DOJ announcement.
S7. Thomas W. Frick, Rodrigo Belo and Rahul Telang, “Incentive Misalignments in Programmatic Advertising: Evidence from a Randomized Field Experiment,” Management Science 69(3), 1665–1686 (2023), doi:10.1287/mnsc.2022.4438. Theory and design §§2–4; results/limits §5 and Tables 3–5; journal pp. 1675–1681 for empirical-strategy caveat, interaction precision, modeled comparison and discussion. Full text read selectively, including methods, tables, robustness and limitations; no underlying data independently replicated.
S8. Randall A. Lewis and Justin M. Rao, “The Unfavorable Economics of Measuring the Returns to Advertising,” Quarterly Journal of Economics 130(4), 1941–1973 (2015), doi:10.1093/qje/qjv023. Abstract-level access only; used narrowly for the measurement-noise warning, with no effect estimate or detailed methods claim.
S9. Sıla Ada, Nadia Abou Nabout and Elea McDonnell Feit, “Context Information Can Increase Revenue in Online Display Advertising Auctions: Evidence from a Policy Change,” Journal of Marketing Research 59(5), 1040–1058 (2022), doi:10.1177/00222437211070219. Relevant design and results: pp. 1040–1051, Tables 1–3; limitations and site-level precision in results/discussion. PDF read for theory, setting, methods, tables/results and stated limitations; confidential exchange microdata and unlinked web appendices unavailable.
Historical facts, management claims, author calculations and teaching assumptions are distinguished. The cutoff excludes TTD's Feb. 15 Q4 earnings release, prepared remarks and call, all later filings and outcomes, and later evidence on browser/product adoption.