The Trade Desk today released Zuma, an update to its Kokai media buying platform built around interface changes, in-platform AI assistance and new measurement tools. Several of the headline capabilities are restricted to closed beta, and the release's central performance figure, a 32 percent average improvement in cost per acquisition, rests on an analysis of 62 campaigns.

The release is labelled as the third-quarter 2026 highlights package from The Trade Desk, and it arrives three weeks after chief executive Jeff Green named the Zuma upgrade first among five priorities for stabilising a business whose shares fell 24 percent on second-quarter results and soft third-quarter guidance. Green had told analysts on August 6 that Zuma would ship later in the month and would concentrate on navigation, workflow and troubleshooting. It has now done so.

According to the company's release page, Zuma introduces an updated user interface, new AI-powered assistance and enhanced measurement tools intended to help traders work more efficiently, understand campaign impact and improve outcomes. The page groups the changes into three categories: improved platform UI, agentic AI and optimization, and enhanced measurement.

Eric Bosco, a product leader at The Trade Desk, described the release in a post published on LinkedIn today. According to Bosco, the work followed a process of collecting and acting on customer feedback. "The objective was not just 'new features' (although there are many!), but a cohesive experience that ties together the vast functionality The Trade Desk provides in a seamless and intuitive way," he wrote. He listed three objectives for the release: faster decision-making through a revised interface and AI-assisted audience expansion, higher return on media spend through AI-driven optimizations, and better evidence of marketing impact through conversion lift management.

What the interface changes actually alter

The first block of changes concerns navigation and comprehension rather than new buying capability.

Refreshed page designs are described as bringing greater clarity to the platform, making it easier to identify trends and locate the areas that require attention. An Applied Settings view surfaces targeting parameters in a single place, with the stated purpose of exposing pacing and performance opportunities that would otherwise sit buried across separate configuration screens.

A third interface change, updates to bulk editing, is marked as coming soon rather than available. According to the release page, the updated bulk editing will allow the potential impact of changes to be previewed before they are applied, so that campaign strategy can be adjusted at scale with greater confidence. No date is attached to that capability.

Bulk editing has been a recurring point of friction on Kokai. When The Trade Desk shipped a set of platform enhancements aimed at faster campaign management in September 2025, industry feedback at the time indicated that bulk upload functionality had been reduced during the migration, with certain placement settings no longer editable in bulk. A preview-before-apply model addresses a different complaint, the risk of large simultaneous edits, but it does not by itself restore the scope of what can be edited.

Agentic AI moves from a separate product into the platform shell

The second block is where the release concentrates its ambition, and also where the availability caveats cluster.

The Koa AI assistant is described as an AI-powered companion inside the platform that helps locate information, uncover insights and provide on-demand support. Behind it, according to the release page, Koa coordinates a growing set of specialised AI agents, each supporting a different part of the media buying workflow, including campaign creation, audience building, troubleshooting and performance analysis. The assistant is in closed beta.

That architecture is a continuation rather than a departure. The Trade Desk shipped Koa Agents in April 2026 as its first in-platform AI agent, with Stagwell as pilot partner, and published the Open Agentic Kit as a proposed interoperability standard on the same day. Reporting in May documented an agent workflow connecting through a partner language modelto ingest a media plan, reformat it into a platform-compatible template and build the campaign. Earlier, in March, PPC Land documented the company's closed testing of Anthropic's Claude for campaign creation. What Zuma adds is a conversational surface that sits above those agents and routes requests to them.

Koa frequency features, also in closed beta, use AI to adjust frequency settings dynamically against campaign objectives. The stated aim is reducing manual guesswork in delivery management.

Koa Optimizations, the existing bid and inventory prioritisation layer, is described as more effective through the combination of real-time performance signals with historical insights. This is the component carrying the 32 percent figure. It is not marked as beta.

Audience Unlimited receives a simplification of access to third-party segments, paired with AI-powered expansion. That product launched in September 2025 with a tiered structure charging 3.3 percent and 4.4 percent of impression costs in Control Mode, and bundling data access at no incremental cost in Performance Mode. It has since absorbed additional inventory of signal, including location data from Adsquare in June 2026. Green named the ramp of Audience Unlimited among his stabilisation priorities on the August earnings call. Zuma does not change its pricing, and the release page does not mention pricing at all.

The 32 percent claim and what sits underneath it

The performance number is the most quotable element of the release, and it carries two layers of qualification that deserve separation.

According to the release page, a global platform analysis showed a 32 percent CPA improvement on average with Koa Optimizations enabled. The source footnote states that the figure comes from The Trade Desk platform data, from campaigns comparing the upgraded model with the previous model, with a sample of 62, and that results are statistically significant at p below 0.05.

Two things follow from that footnote. The comparison is between model versions, not between campaigns using Koa and campaigns not using it, which means the number measures the increment from the upgrade rather than the value of the optimisation layer as a whole. And the sample is 62 campaigns, drawn from a platform that processes advertising impressions at a rate the company has historically described in the millions per second.

Bosco's LinkedIn description of the same statistic differs in framing. His post attributes a "32 percent average improvement in CPAs, at scale" to new AI-driven optimizations. The release page attributes it specifically to Koa Optimizations, and the footnote sets the evidence base at 62 campaigns. A 62-campaign model comparison and a claim of improvement at scale are not the same assertion, and the discrepancy between the two descriptions of the same finding is worth noting for anyone evaluating the release against a media plan. The release page also carries a general disclaimer stating that the information is provided for background, is not a representation or guarantee of future availability, and that timelines are subject to change.

Vendor-supplied performance statistics of this kind are not independently verifiable from outside the platform. No methodology document, campaign selection criteria or confidence interval accompanies the figure.

Measurement gets the least attention and may matter most

The third block covers reporting, and it is the shortest section of the release page.

Conversion lift enhancements, in closed beta, are described as simplifying the process of launching lift studies while adding reporting dimensions intended to inform future campaign planning. An improved reporting homepage directs users to frequently used reports, saved templates and generated reports. A new report builder is presented as offering greater flexibility and transparency in customising reports across available dimensions.

The direction is consistent with where measurement spending has moved across the wider market. Google reduced its incrementality testing budget threshold to 5,000 dollars in November 2025, widening access to causal measurement for smaller advertisers. Meta has pushed advertisers toward hybrid incrementality frameworks calibrated by experiment. Independent vendors have moved in the same direction, with Jamloop opening household-level holdout methodology in its self-serve platform in July 2026 and Nexxen building ghost bidding incrementality into its DSP in March.

Reducing setup friction on lift studies is therefore a competitive necessity rather than a differentiator. Green identified advancing the measurement product as a separate stabilisation priority alongside Zuma on the August 6 call, which suggests the company does not consider this release the end of that work.

Closed beta is the operative constraint

Four of the named capabilities carry a beta or coming-soon label: the Koa AI assistant, Koa frequency features, conversion lift enhancements and bulk editing updates. The release page states plainly that several product highlights are currently in closed beta, that initial capabilities may vary, and that additional functionality is planned over time.

No enrolment criteria, participant counts or general availability dates appear anywhere in the materials. For a media buyer reading the release as a purchasing signal, that gap is the difference between a platform that has shipped a conversational AI assistant and one that has announced it. The distinction has practical consequences for agencies planning fourth-quarter workflows, since capabilities in closed beta cannot be assumed into a resourcing model.

Usability as a recurring theme rather than a new one

The framing of Zuma as a usability release places it in a lineage.

Kokai launched in June 2023 as the successor to Solimar. The migration encountered client resistance and ran longer than planned, and by August 2025 the company was restricting new campaign creation in the legacy interface. The periodic table, Kokai's signature visualisation, was partially retired in September 2025 after sustained criticism from programmatic professionals, with slower-than-expected adoption cited as a contributor to revenue underperformance. Platform difficulties also featured in coverage of executive departures in March 2025.

The quarterly release cadence itself is recent. The second-quarter 2026 Kokai package in May added granular per-component control of Koa Optimizations, connected television pause ads in open beta, Sincera inventory signals and Deal Desk upgrades. Zuma follows the same packaging logic, with the difference that its unifying theme is the shell of the product rather than its buying capabilities.

The commercial backdrop

Product releases at The Trade Desk are now read against a deteriorating financial line.

The company reported second-quarter 2026 revenue of 715 million dollars on August 6, up 3 percent year over year, with adjusted EBITDA of 241 million dollars at a 34 percent margin and third-quarter guidance of at least 650 million dollars. That guidance implies a revenue decline. Shares fell 24.22 percent after hours. The first quarter had beaten consensus at 689 million dollars and still drew downgrades. Leadership has churned, with the company naming its fourth chief financial officer in 14 months in June.

Competitive pressure is documented. Guideline data discussed in May 2026 showed four demand-side platforms holding roughly 85 percent of global programmatic spend, with Amazon moving from under 10 percent to just under 20 percent share in about fifteen months. Amazon DSP has continued to add supply and capability through 2026, including publisher-hosted creatives for Google Ad Manager programmatic guaranteed deals on August 10. On the agentic side, StackAdapt disclosed more than 15,000 internal AI-assisted workflows per week with its Ivy Studio launch in July, a usage disclosure of a kind The Trade Desk has not published for Koa Agents. Green also used the August earnings call to describe rivals wrapping guaranteed deals in agentic AI at low fee levels.

Against that background, a release aimed at making the platform easier to use is a defensive move as much as a product one. Bosco's post closes on the point directly: "With Zuma, we're making it easier for brands and advertisers to get more value from their ad budgets every day. And this is just the beginning."

Why this matters for media buyers

Three practical consequences follow from the release as documented.

The first is that the trader-facing surface of Kokai changes, and training material, standard operating procedures and quality assurance checklists built against the previous interface will need revision. Interface releases carry a hidden operational cost that vendor materials rarely quantify, and this one arrives mid-quarter.

The second is that the conversational assistant, if it moves out of closed beta, alters where campaign knowledge sits. A system that answers troubleshooting questions inside the platform reduces dependence on account teams and internal specialists, and it also creates a new dependency on the accuracy of the assistant's answers. The release materials do not describe how Koa's outputs are validated or logged.

The third concerns the 32 percent figure. Agencies negotiating fee structures or building performance forecasts on the strength of AI optimisation claims are working with a number derived from a 62-campaign model comparison and reported without a methodology document. That is a materially different evidentiary basis from a platform-wide holdout test, and the distinction survives the marketing language in which it is presented.

Timeline

Summary

Who: The Trade Desk, the independent demand-side platform headquartered in Ventura, California, with the release described publicly by Eric Bosco, a product leader at the company, in a post on LinkedIn.

What: Zuma, presented as the next phase of the Kokai platform and published as the third-quarter 2026 release highlights. It covers refreshed page designs, an Applied Settings view, bulk editing updates listed as coming soon, a Koa AI assistant coordinating specialised agents in closed beta, Koa frequency features in closed beta, an upgraded Koa Optimizations model credited with a 32 percent average CPA improvement across 62 campaigns at p below 0.05, simplified Audience Unlimited access with AI-powered expansion, conversion lift enhancements in closed beta, an improved reporting homepage and a new report builder.

When: Released today, August 27, 2026, three weeks after the August 6 second-quarter earnings call at which the release was previewed. No general availability dates are given for the closed beta capabilities or for the bulk editing updates.

Where: Inside Kokai, the platform used by The Trade Desk's advertiser and agency clients globally.

Why: The release addresses usability and workflow complaints that have followed Kokai since its 2023 launch, and it lands as the company works to arrest a revenue slowdown that produced 3 percent growth in the second quarter of 2026 and guidance implying a decline in the third. Green identified the Zuma upgrade as the first of five priorities for stabilising the business on the August 6 call.