Third-quarter earnings call, November 9, 2026: revenue beats on a low bar, but the underlying trend and Q4 guide stay soft

The Trade Desk reported Q3 2026 revenue of $736 million, down 0.5% year-over-year and meaningfully ahead of the outlook management gave in August. The headline flatters the underlying picture. CFO Nate Olmstead disclosed that results included a benefit of approximately 3 percentage points from business with a strategic partner that "we do not expect to recur." Political advertising from the U.S. midterms was a mid-single-digit percentage of revenue. Adjusted EBITDA was $266 million, a 36% margin, also ahead of plan. Adjusted EPS was $0.33, GAAP EPS was $0.08, and free cash flow was $153 million.

Management argued that the business still beat internal expectations even excluding the one-time item. But CEO Jeff Green was candid about the level of growth: "this level of growth is, of course, well below the level of growth that we expect from ourselves in the future."

Q4 guide implies a weak seasonal quarter once political is stripped out

The Q4 outlook calls for revenue of at least $755 million and adjusted EBITDA of at least $280 million. Political should be "a little bit higher" in Q4 than in Q3, so the guide leans heavily on a non-recurring, cyclical tailwind. Oppenheimer's Jason Helfstein calculated that Q3 was down roughly 8% excluding political and the one-time item, while the Q4 guide implies roughly minus 16% on the same basis, against a comparison about 3 points easier. That is an 8-point deceleration, and management did not dispute the framing.

Olmstead attributed the caution to a challenging market and to "some large global advertisers out there that continue to be under external pressure." He said the company is not expecting the normal Q3-to-Q4 seasonal uplift in what is typically its largest quarter. Excluding adjustments, some sequential growth is still expected, "but not at the level that we've seen historically." He also gave a clear warning for Q1: in election years, the sequential decline from Q4 to Q1 has historically been about 25% on an absolute basis. Investors should expect a difficult first-quarter comparison and a visible political air pocket.

On verticals, medical health, personal finance, travel and political were strongest. CPG and auto were called out as soft, with higher inflation and interest rates pressuring auto financing. Olmstead acknowledged that "there's always an element of execution as well." By channel, video including CTV was a low-50s percentage of the business, mobile a high-20s share, display a low double-digit share, and audio about 7%, with audio growing much faster than any other channel.

Dave Pickles returns in an expanded role

The most notable governance development was the announcement that co-founder Dave Pickles, who stepped down as CTO in September 2023 and has served as an adviser since, has agreed to "substantially expand his role once again." Green described him as the architect of the platform and said that bringing his "technical judgment and deep understanding of our platform more directly" into product work will help shape the next generation of ad tech. The company did not specify his title, reporting line or time commitment. Susquehanna's Shyam Patil called the move "another very encouraging step." Green said the pairing rests on "15-plus years of trust that he and I have working together full time."

Cost reset: 15% workforce cut, more than $100 million of 2027 savings

The September reduction of about 15% of the global workforce produced $44 million of restructuring charges in Q3 and is expected to deliver more than $100 million of gross savings in 2027. Olmstead stressed that the objective is not simply margin maximization, and that some savings will be reinvested in the platform, AI, measurement, data and key customer relationships. Management expects expense growth to "moderate meaningfully" in 2027 versus 2026, and said it is looking beyond headcount to platform operations, software, outside services and facilities. Full 2027 operating plans will come in February.

Platform operations expense rose 23% year-over-year, a pressure point. Olmstead cited higher query volumes and greater AI and machine learning use, transition costs from rebalancing owned data center capacity against third-party hosting that are arriving ahead of the efficiencies, and an accounting shift in which certain costs previously netted against revenue are now booked in platform operations. The last factor inflates the year-over-year comparison without a matching rise in economic cost, though it also means the reported cost base is less comparable to history.

On the recent option repricing, Olmstead said no new shares were issued, so there is no added dilution, only a modest increase in stock-based compensation. The company repurchased no stock in Q3 and raised its total available authorization to $1 billion, including $269 million remaining before the increase. It ended the quarter with about $1.6 billion in cash and short-term investments.

Take rate: no change planned, but a door left open

Green said he was "hoping somebody would ask" about pricing, which he said has become a topic for "the peanut gallery." He noted that the take rate has stayed within 1 to 1.5 points of its average across 10 public years, rising about half the time and falling about half the time. His defense is that The Trade Desk is the most transparent platform and therefore invites a debate others avoid by burying fees in media costs. "When you start talking about true cost, it becomes really easy to see that The Trade Desk is not the most expensive platform in the space. Of course, the sticker price, in some cases, is."

He did not rule out a change: "if it's possible for us to lower our take rate and make up more growth, of course, we're open to that." But he said most clients are not truly concerned about cost and that transparency is often used as negotiating leverage. He expects something close to the current level to persist. Asked about outcome-based pricing, Green was dismissive of the model used elsewhere, in which platforms ask for data and money and report results on their own metrics, leaving clients "renting your own data back."

Open measurement launches this quarter, with a growth tailwind promised for 2027

Green called open measurement the company's most exciting product, a measurement framework and marketplace intended to move beyond last-click and last-view attribution. His argument is that last-touch models structurally underfund branding, because "the heavy lifting of branding is done in the premium media," including CTV, sports and news. He positioned The Trade Desk as advantaged because, unlike Amazon and Google, it does not own media: "we are not grading our own homework." The product will use real transaction data and incrementality, extending attribution to advertisers without sophisticated measurement today. Formal launch is expected later in Q4, and Green said it should "improve our growth next year and provide a tailwind for years to come." No pricing, adoption targets or revenue sizing were provided, so the claim is unproven.

Green also said premium publishers are enthusiastic because a third-party validator may show their inventory is worth its price. That links to a change in supply strategy. Where the old mantra was to "go fast and be first," the company now prioritizes the "right access," meaning inventory with enough metadata to enable price discovery. Green said some platforms offer no such metadata and The Trade Desk does not want to facilitate that. He expects LLM inventory to be valuable and expensive, said the company is "okay to be patient" on it, and conceded that "many of our customers are not growing" in this environment.

Zuma and agentic AI: early efficiency data, still in beta

Zuma, launched in Q3 as an upgrade to the Kokai platform, improves navigation, workflows and troubleshooting and adds more AI to the interface. The company disclosed that median active time declined 9% for campaign creation, 15% for ad group creation and 9% for creative uploads. Koa optimizations are enabled on 22% more strategies. In A/B tests across 62 campaigns, the upgraded Koa CPA model improved CPA by a median of 21% with no extra setup.

A campaign creation agent in beta reads a media plan and builds campaign structure, ad groups, budgets and flights in minutes, work that can take hours. Using MCP, Koa can pass audience objectives to data providers' AI systems, which respond with custom segments that Koa uses only if it expects a performance gain. The company has expanded its Stagwell collaboration into agentic workflows for planning, activation and optimization. Green framed agentic AI as "an API that can reason" layered over a platform with "quadrillions of permutations" per campaign. The productivity implications for agency headcount, and thus for the trading-desk labor that has historically operated the platform, were noted but not quantified.

Audience Unlimited, the third-party data product, has been tested by more than 100 advertisers with several hundred campaigns live. Recent campaigns showed CPA improvements above 30% and CPM reductions of roughly 14%. These are selected early results rather than a cohort average.

Commercial execution: joint business plans, international and new clients

Joint business plans are up more than 30% year-over-year through Q3, including growth above 20% in the quarter, with more than 100 in the pipeline. Customers under JBPs are growing meaningfully faster than those on traditional master services agreements, although self-selection likely contributes. Green cited a financial services client where engagement with individual business lines produced millions of dollars of incremental spend from one line alone, illustrating the shift of buying decisions toward procurement and the C-suite.

APAC is up about 30% year-over-year in the first 9 months, with China up 45% in Q3. EMEA is up more than 20% over the same period. The growth team, focused on new and middle-market clients, increased its client count by more than 200% year-over-year, including 23 new clients in Q3, though the base is small.

On whether to push into the mid-market or acquire, Green said the company is "looking more now than we historically have" at M&A, with standards "higher than ever," and does not want to make "our 17th priority our #1 priority simply because we bought a company." He argued that about half of the addressable market sits with the top 1,000 advertisers.

Assessment

The call offers evidence of better execution: a beat, stronger JBP and international growth, and a cost base being cut. But the beat was supported by a one-time partner benefit and political, the Q4 guide implies a sharp ex-political slowdown, and the Q1 election-year step-down is a known headwind. Open measurement and agentic tools carry the 2027 recovery case, yet neither has been monetized or sized. Pickles' return is a credibility signal for product, though its operating impact is undefined. Investors will likely want evidence of underlying growth, not just cost discipline, before treating Q3 as an inflection.