Durability
Durability
The Trade Desk has grown revenue in each of its ten reported fiscal years, so the framework's structural-decline disqualifier is not triggered. But none of the five sources the framework accepts as conviction — market structure, regulatory barriers, capital intensity, essentialness, operating history — applies here on the company's own filed description of itself. What protects year-10 cash flow is execution and a 21.6% take rate; a top-two customer contested the second in 2026, and Amazon is contesting the first.
The conviction sources, one by one
The framework grants conviction from five specific places. Each is graded below against The Trade Desk's own filings rather than against the company's reputation.
Sources: FY2025 Annual Report (Form 10-K) — Our Competition and Our Industry [1], [2]; Our Clients [3]; Executive Summary [4]; consolidated balance sheet [5]; risk factors [6]; Note 1 Nature of Operations [7]; 2025 capex, Consolidated Statements of Cash Flows [8]; cookie and identifier restrictions [9] and competitors holding first-party data and walled-garden inventory [10]; NASDAQ listing date [11].
Two of those entries deserve their exact wording, because the company chose it. On competition: "Our industry is highly competitive and fragmented. We compete with other demand-side platform providers, some of which are smaller, privately held companies and others are divisions of large, well-established companies such as Google and Amazon" [12]. On switching costs: client master services agreements "do not contain any material commitments on behalf of clients" and are "terminable at any time upon 60 days' notice by either party" [13], with "limited cost and difficulty to moving their media spend to our competitors" [14].
What the company does have is real and should not be understated: 430 directly integrated exchanges, publishers and supply-side platforms and more than 370 third-party data vendors [15], a retail-data marketplace management says covers more than 80% of sales from top U.S. retailers against Amazon's under-15% share of U.S. retail spend [16], a client retention rate the FY2024 filing puts above 95% in each of the preceding eleven years [17], and ten straight years of revenue growth. That is a strong competitive position built by operating well. The framework's own construction is that operating well is not a year-10 protection, because it must be re-earned every year against whoever shows up.
The threats, quantified
The take rate is doing the growing
Revenue grew 18% in 2025. Gross spend — the actual volume of advertiser money routed through the platform — grew 11% [18]. The gap is take rate.
Take rate = revenue divided by gross spend; derived from figures as filed. Sources: FY2025 Form 10-K, Executive Summary [19]; FY2024 Form 10-K, Executive Summary [20]; FY2023 Form 10-K, Executive Summary [21].
Held at the 2024 take rate of 20.30%, 2025 gross spend of $13,395m would have produced revenue of $2,719m. Reported revenue was $2,896m. So $177m of the $451m revenue increase — 39% of it — came from charging more per dollar routed, not from routing more dollars [22].
The company's own risk factor names the long-run direction of that variable, and it is the other way: "As the market for programmatic buying for advertising matures, growth in spend may outpace growth in our revenue due to a number of factors, including pricing competition, volume discounts and shifts in media, client and channel mix" [23]. Viant's chief executive, describing the same market from inside it, put a number on the competing offer: Amazon "are really good at subsidizing businesses in the near term — you are seeing that with their 1% fees that they have been out in the marketplace" [24].
The sensitivity is arithmetic. On 2025 gross spend, each 100 basis points of take rate is $134m of revenue. Reverting to the 20.30% of 2022–2024 costs $177m of revenue; at the FY2025 adjusted-EBITDA margin of 41.3% ($1,196m on $2,896m) [25], that is roughly $73m of cash profit — about a quarter of 2025 adjusted free cash flow. A move to 15% takes $887m of revenue, nearly triple adjusted free cash flow. Nothing structural stops that move; the 60-day termination right is on the client's side of the contract.
Amazon: two incompatible descriptions of the same market
In November 2025 the chief executive told analysts that roughly 97% to 99% of Amazon's advertising effort is monetising owned-and-operated inventory, that Amazon's demand-side platform is "definitely a lower priority," and: "Looking ahead to the next ten years, I doubt Amazon will develop a DSP as we understand it" [26].
In the same quarter, Amazon's chief executive told his own analysts the opposite: "We have addressed customer feedback over the past 20 months and closed key gaps, making our DSP fully featured. Our partnership with Roku provides the largest connected TV presence in the U.S. Furthermore, we have added integration opportunities with ad inventory from Netflix, Spotify, and SiriusXM for our DSP customers" [27]. One quarter earlier Amazon described the Roku arrangement in terms that matter commercially: advertiser access to "80 million connected TV households — the largest authenticated connected TV footprint in the U.S. — exclusively through Amazon DSP," alongside an integration with Disney's real-time ad exchange [28].
The resource asymmetry behind that is filed: Amazon's advertising services revenue was $68.6bn in 2025, up from $56.2bn and $46.9bn in the two prior years [29]. The Trade Desk's entire revenue is $2.9bn and its entire gross spend $13.4bn. A competitor 24 times larger on the revenue line can price a demand-side platform at 1% for as long as it wants to.
Connected television and other video was a low-50s percentage of revenue in Q1 2026 [30], roughly $1.5bn annualised. If exclusive or owned-platform supply paths of the Roku kind take a quarter of that pool over ten years, that is about $375m of revenue and, at the 41.3% margin, roughly $155m of cash profit — more than half of 2025 adjusted free cash flow. This is a scenario, not a forecast; what makes it a live one rather than a hypothetical is that the first such arrangement has already been announced by the counterparty, in its own words, with the word "exclusively" in it.
Two customers, one withdrawn recommendation
If client relationships were aggregated at the holding-company level, two holding companies would each have represented more than 10% of gross billings in 2025, against one in 2024 [31]. At year-end 2025 two clients each accounted for at least 10% and together 30% of consolidated receivables [32]. Publicis Groupe was named in the FY2023 filing as the one holding company above that 10% line in both 2023 and 2022 [33].
That relationship broke in public. Ad Age reported in mid-March 2026, and Digiday and MediaPost followed, that Publicis stopped recommending The Trade Desk to clients after an audit alleging discrepancies in fees, consent and cost transparency; the two sides issued a joint statement in mid-June 2026 saying the matter was settled and Publicis was recommending the platform again, without disclosing terms. Asked about it on the Q1 2026 call, the chief executive said only that "our negotiations are ongoing" and that the coverage had been "overdramatized" [34]. Three months of a top-two customer's recommendation withdrawn, over fee transparency, is the concentration risk and the take-rate risk arriving as the same event. Losing one such holding company would remove at least 10% of gross billings — using 2025 gross spend of $13.4bn as the proxy for gross billings, at least $1.34bn of spend and roughly $290m of revenue at the 21.6% take rate, about $120m of cash profit at the FY2025 margin, or 40% of adjusted free cash flow.
The open-internet pool
Google Network revenue — Google's advertising on third-party publisher properties, structurally the same pool The Trade Desk buys in — was $31.31bn in 2023, $30.36bn in 2024 and $29.79bn in 2025 [35]. Three consecutive annual declines, 4.9% cumulatively, while Google's total advertising grew. That series does not separate a shrinking open web from Google losing share of it, and The Trade Desk's own gross spend grew across the same period — so it is a caution, not a finding. It is the only recurring quantitative disclosure in the peer set that measures the open web directly, and it points down.
The disqualifier check
The framework's one automatic exclusion on this pillar is revenue declining at a high-single-digit rate for three consecutive fiscal years after long existence.
Source: fit_features.revenue_trajectory, derived from company filings as reported; FY2025 figure also filed at FY2025 Form 10-K, Executive Summary [36].
Source: fit_features.revenue_trajectory, year-on-year percentages as computed in the deterministic feature file from filed revenue.
fit_features.revenue_trajectory records consecutive_decline_years: 0 and three_year_hsd_decline: false. Revenue has risen in every one of the ten fiscal years on record, a compound rate of 34.4% from $203m to $2,896m. The disqualifier is not triggered, and the structural-decline exclusion is checked and absent at the company level.
What the same series shows is a decelerating slope: 25.6% in 2024, 18.5% in 2025, 11.8% in Q1 2026 ($689m against $616m), and a Q2 2026 guide of "at least $750m" against $694m a year earlier — 8.1% if met at the floor [37]. Deceleration is not decline, and the two should not be blurred. Growth of 8% is still growth.
FCF consistency
fit_features.fcf_stability returns an empty series. The feature file records the reason: adjusted free cash flow is not_computable because stock-based compensation is absent from the structured cash-flow feed for every year from FY2016 to FY2025, so no five consecutive adjusted years exist. The profile's rolling five-year average adjusted-FCF series therefore cannot be reported from the deterministic file, and that is recorded as a data gap.
The filings themselves do disclose both components. The series below is built from the cash-flow statements as filed, using the profile's definition, and is shown as filed arithmetic rather than as a substitute for the feature file.
Adjusted FCF = reported free cash flow minus stock-based compensation minus the trailing five-year average of acquisition spend; components as filed. Sources: FY2025 Form 10-K, Consolidated Statements of Cash Flows [38]; FY2023 Form 10-K [39]; FY2021 Form 10-K [40].
The acquisition leg is negligible and that is worth stating precisely: the only two purchases in the record are $13.3m in 2021 [41] and $4.4m in 2025 [42], a five-year average of $3.5m in the FY2025 window. Stock compensation is the whole of the adjustment: $490.6m in 2025 against reported free cash flow of $795.7m, so 62% of the reported cash flow is paid in shares.
Rolling five-year average of the adjusted-FCF series above; derived from filed cash-flow statements [43], [44], [45]. Only three complete windows exist; stock compensation before FY2019 is not in this corpus.
Read against the framework's test — is the rolling five-year average stable, or genuinely unpredictable — the series is neither volatile nor cyclical. It rises monotonically, $33m to $73m to $90m, and the two negative years, 2021 and 2022, are not an underwriting cycle. They are the years the stock-compensation line stepped from $112m to $337m to $499m. That is a discretionary compensation decision, not a business-model rhythm with a five-to-eight-year cadence, so the framework's tolerance for periodic negative episodes does not apply. Since 2022 the line has been flat in dollars — $499m, $492m, $495m, $491m — while revenue grew 84%, taking stock compensation from 31.6% of revenue to 16.9%. That flattening, not cash-flow growth, is what turned adjusted free cash flow positive.
The base is also small and young. Three complete windows is a short record on which to call anything stable, and the latest window's $90m average sits against a market capitalisation of $8.53bn (fit_features.market_cap, $17.29 close on 24 July 2026). The yield arithmetic belongs to Yield; what it means here is that today's adjusted free cash flow is a low bar for year 10 to clear, which cuts in favour of the gate, not against it.
The year-10 case
The strongest case that year-10 revenue and adjusted free cash flow are higher. Ten consecutive years of revenue growth through a pandemic and the 2022 advertising recession; revenue 14.3 times its 2016 level. The pool is growing — the filing puts digital advertising above $700bn and total advertising past $1tn, with the linear-to-connected-television shift early [46]. Adjusted free cash flow is compounding hard off a low base, from minus $37m in 2022 to $302m in 2025, and the operating leverage that produced it — stock compensation flat in dollars against 84% revenue growth — has further to run. The balance sheet cannot force the outcome: $658m of cash and $645m of short-term investments at year-end 2025 [47], about $1.4bn of cash and short-term investments at the end of Q1 2026 [48], and no debt drawn on the $450m revolving credit facility at year-end 2025 [49]. And consensus underwrites growth, not decline: fit_features.consensus_forward_yield carries mean free-cash-flow estimates rising from $780m in FY2026 to $1,098m in FY2029. Even a decade of 5% growth from here leaves year-10 revenue near $4.7bn.
The strongest doubt. Not one of the five conviction sources applies, on the company's own filed description of its industry, its contracts and its assets — so year-10 protection rests on out-executing Google and Amazon in a market the company itself calls fragmented, and the framework does not accept execution as a moat. The volume measure is decelerating faster than the revenue measure: gross spend grew 11% in 2025 against 25% in 2024, and 39% of the year's revenue increase came from a take rate that rose to 21.6% while the company's own risk factor names compression as the mature-market direction and a competitor with $68.6bn of advertising revenue [50] is described by a rival platform as being in the market at 1% fees [51]. Two holding companies each control more than 10% of gross billings, and one of them withdrew its recommendation for three months in 2026 over exactly that fee question. Amazon has told its own shareholders that its demand-side platform is now "fully featured" and that access to Roku's 80 million connected-television households runs "exclusively" through it — in the same quarter The Trade Desk's chief executive told his shareholders he doubted Amazon would build such a platform at all in the next ten years. Both statements cannot be right in year 10.
The read. The framework's gate is not met. The doubt is not that revenue declines — the disqualifier is checked and absent, and the base case on this evidence is continued growth — but that year-10 revenue and adjusted free cash flow cannot be underwritten with very high conviction when the protection is execution plus a price that a customer has already challenged and a far larger competitor can undercut to zero. Under the framework's own construction, any proper doubt resolves the same way, and this doubt is specific, cited and quantified rather than atmospheric.
What would change it, in order of how much weight each carries: gross-spend growth running at or above revenue growth for four consecutive quarters, which would show volume rather than price doing the work; the two holding-company concentrations falling below 10% each without a revenue step-down; and a multi-year record of premium connected-television supply staying biddable on independent platforms after the Roku arrangement. The first is measurable in the next two filings.