The Trade Desk's 0.5% decline in third-quarter revenue would have been closer to 3.5% without business from a single strategic partner that the company does not expect to repeat, chief financial officer Nate Olmstead told analysts on November 9, 2026. Speaking on the conference call that followed the results, Olmstead said the partner added about 3 percentage points to revenue, that US midterm political advertising made up a mid-single-digit percentage of the business, and that first-quarter revenue has historically fallen about 25% from the fourth quarter after election years. Chief executive Jeff Green used the call to defend the company's fees, to promise an "open measurement" product before the end of the year, and to argue that AI chatbots have brought competition to what he called a monopoly in search.

In Short

On its results call on November 9, The Trade Desk said a one-time deal with an unnamed partner added about 3 points to a quarter in which revenue slipped 0.5%, and that midterm election ads were a mid-single-digit share of sales. Both boosts fade in 2027, and the company says first-quarter revenue has usually dropped about 25% from the fourth quarter after an election, which matters to the advertisers, agencies and publishers that rely on one of the biggest independent ad-buying platforms. The next changes are a measurement product due before year-end, which the company pitches against Amazon and Google measuring their own ads, and more AI agents inside its buying software.

What sat inside a flat quarter

PPC Land covered the release, the 67% fall in net income and the $1 billion buyback authorisation on the day. It left two questions open: how much political spending had contributed, and whether anything unusual sat inside the revenue figure. The call, which began at 2:00 p.m. Pacific time, answered both.

"Our third quarter revenue was approximately $736 million, meaningfully ahead of the outlook we provided in August," Green said in his opening remarks. "While these results are above what we expected for this quarter, this level of growth is, of course, well below the level of growth that we expect from ourselves in the future."

Olmstead then supplied the detail the release had not. "In addition, our Q3 results included a benefit to revenue totaling approximately 3 percentage points from business with a strategic partner that we do not expect to recur," he said. He did not name the partner or describe the business, and no analyst asked. Three percentage points of the $739 million reported a year earlier is roughly $22 million, by PPC Land's arithmetic. "Importantly, even excluding that one-time benefit, the underlying business still performed meaningfully better than we anticipated entering the quarter," Olmstead added.

Political advertising was the second piece. "It represented a mid-single-digit percentage of our business in Q3, and we expect it should be a little bit higher in Q4 than it was in Q3," Olmstead said in reply to Justin Patterson of KeyBanc. Read as 4% to 6%, that is somewhere between about $29 million and $44 million of third-quarter revenue, again a PPC Land calculation. The third quarter of 2025 offered no comparable lift: a year ago, The Trade Desk reported growth of 18%, or 22% excluding political spending from the prior year, because 2024 had been a presidential election year.

Jason Helfstein of Oppenheimer put the two pieces together. By his estimate, third-quarter revenue was down about 8% once political spending and the one-time partner business were removed, and fourth-quarter guidance implied a decline of about 16% on the same basis, against a comparison he described as 3 points easier. That, he said, amounted to an 8-point slowdown. Olmstead did not dispute the figures. PPC Land's own rough arithmetic, using the midpoint of the political range, lands at a similar decline of about 8% for the third quarter.

Verticals, channels and regions

Olmstead named the categories that grew fastest. "We saw particular demand strength in medical health and the political election business, which continued to grow cycle on cycle as expected," he said, adding personal finance and travel to the list of the strongest year-over-year performers.

Asked by Mark Mahaney of Evercore which verticals were weak, he pointed to the same two categories that were under pressure in August, when Green said consumer packaged goods and automotive together accounted for around 25% of the business. "On the verticals, I would probably call out CPG and auto as a couple where we have seen some pressure," Olmstead said. "I think the higher inflation, higher interest rates impacts auto financing and things like that."

By channel, the mix was as follows:

  • Video, including connected TV: "a low 50s percent share of our business in Q3."
  • Mobile: "a high 20s percent share of the business during the quarter."
  • Display: "a low double-digit share."
  • Audio: "around 7% of the business," and it "posted much stronger year-over-year growth than any other channel."

The geographic picture came from Green. "Outside the U.S., we are seeing strong growth from our APAC business, which is up approximately 30% year-over-year in the first nine months, led by China, which grew 45% year-over-year in Q3," he said. EMEA was "up more than 20% year-over-year in the first nine months."

Green gave no figure for North America, which accounted for about 86% of 2025 revenue according to the company's investor presentation. Total revenue rose about 4% over the first nine months. If the international regions kept roughly their 2025 share of 14% and grew by 20% to 30%, they would account for most of that growth, leaving North America close to flat. That is a PPC Land estimate, resting on assumptions the company has not confirmed.

The fourth quarter and the hangover after the vote

The fourth-quarter floor of $755 million is about 2.6% above third-quarter revenue. A year earlier, revenue rose about 14.5% between the same two quarters, from $739 million to $847 million. Both comparisons are PPC Land calculations.

Olmstead explained the gap directly. "So while Q3 performed a little bit better than we expected, we're not assuming that every positive indicator that we see immediately translates into revenue recognition," he said. "There's some large global advertisers out there that continue to be under external pressure." The quarter is normally the company's largest. "We're not expecting the normal seasonal uplift Q3 to Q4 that we've seen in the past. We're expecting some growth."

He also pointed analysts beyond December. "Typically, or I would say very consistently, Q1 is always seasonally lower than Q4, and in election years, the sequential decline has historically been about 25% on an absolute basis from Q4 to Q1," Olmstead said. "As you think about Q1 and you start putting things together, you should definitely be mindful of that historical trend."

The company gave no first-quarter guidance, and the $755 million is a floor rather than a forecast. Applied mechanically to that floor, a 25% decline would put first-quarter 2027 revenue at roughly $566 million, against $689 million in the first quarter of 2026. That is PPC Land's arithmetic, not a company projection.

Green presented the political result as evidence of trust. "This election was our biggest yet," he said, and later: "We had a bigger growth rate in political compared to the last presidential election." He added a caveat: "It is important to note, though, that political will always be cyclical."

Elsewhere in ad tech, the midterms produced similar surprises. Magnite, the sell-side platform, said its political contribution came to about $12 million in the third quarter and would exceed $40 million for 2026, against the $15 million it had projected.

Costs, savings and a change in accounting

Third-quarter operating expenses totalled $643 million, including $44 million of restructuring charges tied to the September cuts. "Excluding stock-based compensation and restructuring charges, Q3 operating expenses were $504 million," Olmstead said. The workforce reduction, which PPC Land reported at about 575 positions, or roughly 15% of staff, is expected to save money from next year. "We currently expect these workforce actions to generate more than $100 million of gross savings in 2027," he said. That is equivalent to about 5% of the $504 million quarterly cost base multiplied by four, PPC Land's calculation.

Not all of it will reach profit. "Our objective is not simply to maximize near-term margins," Olmstead said. In the question-and-answer session, Youssef Squali of Truist Securities asked what lay beyond headcount. "Some of that will be reinvested, some of that will flow through," Olmstead replied, listing other targets: "Obviously, platform operations is one area, software, SaaS, outside services, facilities, really across the board." Overall, he said, "we estimate a meaningful moderation in our expense growth compared with 2026," with a fuller 2027 operating plan due when the company reports in February.

Why platform operations rose 23%

The release had shown platform operations expense up 23% year over year. Olmstead gave three reasons.

"First, we continue to invest in the infrastructure required to support the increasing scale and sophistication of our platform, including higher query volumes and greater use of AI and machine learning," he said. "Second, we are investing to optimize the mix between owned data center capacity and third-party hosting. Some of the costs associated with that transition are occurring ahead of the efficiencies we expect the infrastructure to generate over time."

The third reason is an accounting matter, and it bears on more than costs. "Third, as our platform and value-added services and data evolve, certain costs that historically were recognized as a reduction to revenue are now being recognized as a cost within platform operations expense," Olmstead said. "That affects the year-over-year OpEx comparison, even though it does not represent a comparable increase in the underlying economic cost of operating the platform."

The implication runs in both directions. Costs that were once netted out of revenue now sit in expenses, which inflates the expense comparison. By the same logic, amounts that would previously have been deducted from revenue are now included in it, which flatters the revenue comparison. Olmstead did not quantify the effect on either line, and no analyst asked him to.

Tax, options and buybacks

The release had shown income tax of $68 million on pre-tax income of about $107 million, an effective rate PPC Land calculated at about 64%. Olmstead's explanation was brief. "Income tax expense was $68 million in the third quarter, driven primarily by our pre-tax profitability and the impact of stock-based awards," he said.

Squali also asked about a recent repricing of employee stock options. "In terms of dilution from the option repricing, we actually did not issue any new shares through the option repricing, so there is no additional dilution from that," Olmstead said, while acknowledging "a little bit of additional stock-based compensation." Neither he nor Squali gave the date, the number of options or the new exercise price.

On buybacks, the release had shown $269 million left under the program at both June 30 and September 30. Olmstead confirmed what that implied. "We did not repurchase shares of our Class A common stock during Q3," he said. The new authorisation brings the amount available to $1 billion.

Measurement aimed at the walled gardens

Green devoted the largest part of his prepared remarks to measurement. "The advertising industry has a fundamental measurement problem," he said. "Last click and last view measurement naturally favor advertising that occurs right before a transaction takes place."

His argument was that this pushes budgets away from the media The Trade Desk buys. "Branding will always suffer and be underfunded when last touch attribution is the default, because the heavy lifting of branding is done in the premium media," Green said. He cited a client who had told him earlier in the year: "Often the best and most effective ads for our brand are the most expensive."

The company positioned the product against the walled gardens, naming two of them. "Unlike the walled gardens of Amazon and Google and other user-generated content platforms, we are not grading our own homework," Green said. "We are providing real sales data through real transactions." He added: "We don't own media that we are trying to sell."

On how it will work, Green offered outlines rather than specifications. "For advertisers with first-party data, that means more granular attribution. For advertisers that do not have access to sophisticated measurement today, we intend to make attribution available more broadly," he said. The design, he said, uses incrementality to separate growth that advertising caused from "the growth that would have happened anyway."

The launch date is near. "I'm excited to share more, and we will when we formally launch this open measurement product later this quarter," Green said, which places it before the end of December. In reply to Shyam Patil of Susquehanna, he set expectations for its financial effect: "While open measurement will be implemented this quarter, we expect the product to improve our growth next year and provide a tailwind for years to come."

The call gave no product name, price, list of retail or data partners, or methodology. Nor did Green say how the product relates to the third-party measurement providers already sold through the platform, or whether advertisers will pay for it separately.

Zuma and Koa: the figures behind the agents

Green described Zuma, released in August, as an upgrade to the existing buying platform rather than a replacement. "Zuma upgraded Kokai," he said. He then gave a set of usage figures:

  • "Median active time to complete core setup workflows has declined, including 9% for campaign creation, 15% for ad group creation, and 9% for uploading creatives."
  • "Koa optimizations are now enabled on 22% more strategies than before."
  • "In A/B tests across 62 campaigns, the upgraded Koa CPA model improved CPA by a median of 21% versus the prior model, with no additional setup changes required."

The last figure differs from the one the company used at launch. When Zuma was released on August 27, The Trade Desk cited a 32% average improvement in cost per acquisition across 62 campaigns. The November figure is a median of 21% across the same number of campaigns. An average can sit well above a median when a few campaigns improve sharply, so the two numbers need not conflict. Green did not say whether the two sets of campaigns are the same, or give the spread of results.

A campaign creation agent, still in beta

Green described an agent that sets up campaigns from a media plan. "In early testing, our campaign creation agent read a client's media plan and set up the campaign structure, ad groups, budgets, and flights in a matter of minutes, automating work that was traditionally done manually and can take hours," he said. "We are still in beta, and there's a lot more work to do."

The company is also connecting Koa to outside data companies. "Using MCP, an open standard adapter that allows AI systems to communicate with one another, Koa can share an advertiser's audience objectives with participating data providers," Green said. Those providers' systems return custom segments, and "Koa can then evaluate those options and use a segment only when it believes it is likely to improve performance." He did not name the participating providers.

Stagwell, which PPC Land reported was the pilot agency for Koa Agents when they were introduced in the spring, came up again. "Today, we have extended that collaboration model, incorporating agentic workflows into the planning, activation, and optimization processes that power campaigns for many of the world's largest brands and their agencies," Green said, without naming other agencies.

His description of where this leads has consequences for agency staffing. "Work that has historically required many people and significant manual effort can increasingly be assisted or executed by agents, allowing leaner teams to operate even more sophisticated programs," he said. "Zuma is not a one-time release."

Audience Unlimited after a year

Audience Unlimited, the AI-scored data marketplace product announced on September 29, 2025 with a Control Mode priced at 3.3% and 4.4% of impression costs, received its first adoption figures. "More than 100 advertisers have now tested the product, and several hundred campaigns are live," Green said. "In recent campaigns, advertisers have improved CPA by more than 30% and simultaneously reduced CPMs by approximately 14%."

He gave no sample size, time period or comparison baseline for those results. "Multiple advertisers have also used Audience Unlimited successfully in direct competitive tests," he added, without naming the competitors or the measure of success.

Shweta Khajuria of Wolfe Research asked why the company's take rate pricing model remained the right one. Green said he had been hoping for the question, because the subject had become a topic of conversation "seemingly in the peanut gallery."

His defence rested on history and on transparency. "It's been up about half of those 10 years, and it's been down about half of those 10 years, but all within a point and a half, if you will, of where the average is," Green said of the take rate since the 2016 listing. He made a similar point in August, when he told analysts the take rate had risen in five of the past 10 years and fallen in five. Neither call gave the take rate itself.

Green argued that disclosure invites criticism competitors avoid. "By giving transparency, we can incite debate where others will just bury their costs into media or hide it in some other way," he said. He conceded one point. Asked about cost, Green said The Trade Desk is not the most expensive platform once all costs are compared, then added: "Of course, the sticker price in some cases is." Competitors, he said, had sacrificed objectivity: "we preserve our objectivity, where almost every one of our competitors have burned that to the ground."

He left room for change, with conditions. "If it's possible for us to lower our take rate and make up more growth, of course we're open to that," Green said. "But in most cases, what we found is that people aren't really concerned about the cost. Instead, our transparency can create discussion or debate, or in some cases, just give leverage for a negotiation."

The fee debate has not been confined to analysts this year. Earlier in 2026, Publicis Groupe advised clients to stop transacting on The Trade Desk after an audit by FirmDecisions concluded that the platform had applied its DSP fee to other fees and opted Publicis and some clients into fee-based products without documented authorisation. The Trade Desk disputed the findings. The two companies issued a joint statement on June 12, 2026 resolving the dispute, after which Publicis reinstated The Trade Desk as a recommended DSP for its clients. Neither side disclosed terms. Green did not refer to the episode on the call.

Against outcome-based pricing

Helfstein asked whether the company could move the conversation to outcome-based pricing. Green rejected the model as practised elsewhere: "Most of the companies in advertising who have outcome-based pricing essentially go to clients and say, 'Give me your data and give me your money, and I will just give you results.'" Clients, in his account, end up "essentially renting your own data back" and depending on the platform's own reporting. He did not name those companies.

Premium supply, and the chatbots

Matt Swanson of RBC Capital Markets asked why some new premium inventory had launched with other DSPs first, when The Trade Desk had once been the default. Green said the company had changed its approach. From 2017 until the pandemic, he said, its inventory partnerships team worked to the mantra "Go fast and be first." That ended, he said, "in part because there is so much more supply than there is demand, that there isn't a deep need for more supply."

He was critical of some sellers. "There are a whole bunch of platforms, partly because of that outsized supply to demand, that are not trying to enable price discovery, that don't want to provide any metadata so that an advertiser doesn't know what they're buying," Green said. "We do not want to facilitate that."

He then turned to advertising in AI chatbots. "I do expect that TAM to open up, in part because I do expect their inventory to be really, really valuable, but it's also going to be very expensive, in part because of all the processing power required for them to give great answers," he said. "But we don't want to start off by doing it wrong, and we're okay to be patient."

The context for the question is recent. On September 10, Amazon Ads and OpenAI opened a US pilot that lets select advertisers extend campaigns already running in Amazon DSP into ChatGPT. Green did not mention it. "It is no longer about getting the first access," he said. "It is about getting the right access and enabling price discovery."

He also made an unusually direct remark about clients. "Many of our customers are not growing over the last few years in this environment, and we need to make certain that we are helping them to grow," Green said.

AI, search and the size of the market

Mahaney asked how AI Overviews, AI Mode, chatbots and agents would affect the advertising market. Green began with Google. "I personally think that the most interesting story of this earnings season until recently was Google search growth moving down so much, then of course pointing to the LLMs," he said. "To me, if you zoom out on that, what that means is that there once was a monopoly that controlled all of search, and the LLMs have created competition."

The numbers behind the remark: Alphabet reported on October 28 that Google Search and other revenue grew 12% in the third quarter, down from 17% in the second, and the next day its Class A shares fell 12.6%, their worst day since 2004. Search revenue was still growing. Green's view is that competition will release spending into the open market. "So, you asked the question, will the LLMs and chatbots and all of this unlock TAM? I see no way that it wouldn't," he said.

He described programmatic buying as a natural fit for agents, because the platform's API offers "quadrillions of permutations for every single campaign." In his summary: "If you want to be really simple or reductive about it, agentic AI essentially makes it so you have an API that can reason."

In his prepared remarks, Green put the company's share at "only a small fraction of roughly a $1 trillion global advertising market," and made a point that cuts both ways for a business paid on the media it buys: "Access to inventory is increasingly commoditized." His case is that clients pay for decisioning, not access. "In an AI-driven world, proprietary data becomes even more valuable," he said.

Clients, contracts and acquisitions

Green said joint business plans, the multi-year agreements the company signs with agencies and brands, were the clearest sign of progress. "Through the end of Q3, the number of JBPs was up more than 30% year over year, including growth of more than 20% in the third quarter, representing hundreds of agencies and brands," he said. He did not explain what the second figure measured. He said clients under these plans "continue to grow meaningfully faster than those under traditional MSAs." He added: "We currently have more than 100 JBPs in the pipeline."

The growth rate has slowed through the year. In the first quarter, the company said its JBP count had grown 55% year over year. In the second, it reported JBPs with 217 clients, up 38%. The third-quarter figure is "more than 30%," with no absolute count.

Green offered one example, a large financial services client whose individual business lines the company had courted. "That engagement recently resulted in millions of dollars of incremental spend from one line of business alone," he said. A separate growth team, aimed at new brand clients and lapsed advertisers, "has increased its client count by over 200% year-over-year, including 23 new clients in Q3 alone." He gave no base for the percentage.

He also described a shift in who decides. "Decisions about what advertising to buy that were once made primarily by hands-on keyboard traders are increasingly being made by senior agency leaders, procurement groups at large brands, and senior executives," Green said.

The mid-market and M&A

Vasily Karasyov of Cannonball Research asked whether the company would push into the mid-market, try new business models or buy growth. Green said the company had always started with the largest advertisers and worked down, and noted that "about half the TAM is contained inside of the top 1,000 advertisers, if you will." By TAM he meant the total addressable market.

On acquisitions, he signalled more activity than before. "I will just transparently say, we're looking more now than we historically have, partly because I just think there's a bunch of opportunity in the market that there hasn't been in the past," Green said. "That said, it's also critical to take away that our standards are also higher than ever." He explained why: "I would never want to derail that by making our 17th priority our number one priority, simply because we bought a company."

Pickles, mistakes and focus

Early in his remarks, Green turned to Dave Pickles, the co-founder who stepped down as chief technology officer in September 2023 and whose return to lead a new CTO Council was announced on November 3. "Dave, the architect of the platform and a key driver of The Trade Desk's revenue, has agreed to substantially expand his role once again," Green said. He did not give Pickles' title or reporting line.

Green was also candid about the past year. "We've made mistakes, and when we do, we try to correct them quickly," he said. Answering Patil, he was more specific: "At times, we have been too focused on doing 100 small things instead of 10 big things that can move the needle for our clients."

What the call did not cover

Several subjects that weigh on the business did not come up, either in prepared remarks or from analysts.

Safari. Apple's iOS 27, released on September 14, stopped The Trade Desk serving ads on Safari by blocking adsrvr.org and UIDAPI.com. No one on the call mentioned Apple, Safari or Unified ID 2.0.

Amazon as a DSP rival. Amazon appeared once, as a walled garden in Green's measurement pitch. Its demand-side platform, which Guideline data cited by PPC Land in August showed rising from under 10% to just under 20% of tracked programmatic spend in roughly 15 months, was not discussed. Neither was Walmart.

The partner. The identity of the strategic partner behind the 3-point revenue benefit, and the nature of that business, went unstated.

Spend and fees. As in the release, there was no figure for gross spend on the platform or for the take rate.

Shares and expectations

According to Investing.com, analysts had expected adjusted earnings of $0.25 a share on revenue of $650.89 million; the company reported $0.33 and $736 million. The shares had already risen 15.47% during the regular session on November 9, to $14.93 from $12.93, before the results were published after the market closed. They added a further 2.15% in after-hours trading, to $15.25. Investing.com put the 52-week range at $11.62 to $45.50.

Why this matters for marketers

For agencies and brands that buy through The Trade Desk, the most practical news is the measurement product. Green's argument - that last-click attribution sends money to the bottom of the funnel and to platforms that measure their own ads - is one many brand advertisers already make. Whether open measurement moves budgets depends on details the call did not provide: data sources, method, price, and whether anyone other than the company can check the results. The claim that The Trade Desk is "not grading our own homework" will be tested on that last point, because the company also earns more when clients spend more through it.

The performance figures deserve the same scrutiny. A 32% average became a 21% median across 62 campaigns within ten weeks, and Audience Unlimited's CPA and CPM gains came without a sample size or baseline. They are company statements, not independent findings.

On fees, nothing changed. The take rate stays roughly where it is, with discounts only where growth would make up the difference. Green's concession that the "sticker price" is in some cases the highest, and his remark that transparency gives clients "leverage for a negotiation," describe the pressure buyers and holding companies have applied this year.

For publishers, the end of "Go fast and be first" means new premium supply, AI chatbot inventory included, cannot count on The Trade Desk signing early, and its insistence on metadata and price discovery is a condition of its demand. Stripped of political spending and the one-off partner, that demand fell by about 8% in the third quarter, by Helfstein's reckoning. If Olmstead's historical pattern holds, the first quarter of 2027 will be the first in which neither prop is present, and the first test of whether open measurement, Zuma and the restructuring show up in the numbers on their own.

Timeline

Summary

Who: The Trade Desk, the Ventura, California-based demand-side platform, through chief executive and co-founder Jeff Green and chief financial officer Nate Olmstead, answering analysts from Susquehanna, KeyBanc, Cannonball Research, Wolfe Research, RBC Capital Markets, Truist Securities, Evercore and Oppenheimer. The statements affect agencies, brands, publishers and data providers that work with its Kokai platform.

What: Olmstead said about 3 percentage points of third-quarter revenue came from a strategic partner whose business will not recur, that political advertising was a mid-single-digit share of revenue and will be a little higher in the fourth quarter, and that first-quarter revenue has historically fallen about 25% from the fourth quarter after election years. He attributed part of a 23% rise in platform operations expense to costs previously recorded as a reduction of revenue, and said job cuts will save more than $100 million in 2027. Green announced an open measurement product for launch this quarter, gave new figures for Zuma and Audience Unlimited, defended the take rate, said the company is looking at acquisitions more than before, and argued that AI is breaking up search.

When: The call took place on November 9, 2026, at 2:00 p.m. Pacific time, after third-quarter results were released the same day. Open measurement is due before the end of 2026, and the 2027 operating plan in February 2027.

Where: The call was held for investors and analysts. Green cited growth of about 30% in APAC and more than 20% in EMEA over nine months, with China up 45% in the third quarter; he gave no North American figure.

Why: Green said growth remains "well below the level of growth that we expect from ourselves," and presented measurement, agentic AI and closer client relationships as the route back. Olmstead framed the fourth-quarter guidance as prudent, citing large global advertisers "under external pressure" and weakness in consumer packaged goods and auto.