Tatari Max Uses AI to Optimize Direct-Sold TV Ad Impressions

Tatari is bringing outcome-based optimization to premium, direct-sold TV, extending a model familiar from digital advertising across a broader mix of TV inventory. Its new engine, Tatari Max, selects which impressions to serve to each brand in real time, using measured performance data to guide those decisions. The pitch is that TV buying can make more use of what campaigns have already shown about their audiences and results, rather than treating each placement as an isolated decision. That’s a meaningful shift in emphasis, though the available evidence is still early.

An intelligence layer for direct TV buys

Tatari Max works at the ad-server level, where it applies AI across the bidstream and calculates a proprietary affinity score from performance factors. The aim is to match impressions with campaigns based on expected outcomes. In practical terms, that means the system is meant to help decide which available impression is a better fit for a particular campaign as inventory is being served. The company hasn’t disclosed the scoring formula, so buyers will have to judge the system by its results and the detail Tatari provides about measurement.

Tatari positions the approach as distinct from standard DSP and programmatic optimization, and says it can cover the full TV spectrum, including linear. That distinction matters because direct-sold TV is typically bought through publisher relationships and negotiated campaigns, while programmatic systems are built around automated buying and selling. Bringing real-time selection into direct buys could give advertisers more responsive campaign management without changing the underlying character of the inventory. But the claim depends on how widely the system can make decisions across participating publishers and formats, and on whether those decisions translate into outcomes advertisers care about.

The engine sits on Upstream, Tatari’s system for automating direct media execution. Announced in early February, Upstream launched with NBCUniversal, Skydance Advertising—formerly Paramount Advertising and Warner Bros. Discovery Advertising—Disney, and Tubi. Tatari says the setup gives brands access to premium inventory without programmatic intermediary fees, while avoiding fraud risk and maintaining brand safety. Those are relevant selling points for advertisers weighing automation against control: direct access may reduce some costs and simplify buying, while publisher relationships can give campaigns a clearer path to premium placements. The practical value will depend on the inventory available to each campaign and the terms of each buy.

Data drawn from measured outcomes

Tatari Max is trained on the company’s Foundational TV data set, which Tatari says includes 300 terabytes of response data, 180 million conversion events processed daily, more than $9 billion in managed spend, 1.4 trillion linear impressions, and performance scores for more than 23,000 creatives. The data is aggregated and anonymized. Those figures describe the scale of the company’s accumulated measurement, rather than independently establishing how accurately the engine predicts results for a new campaign. Still, a large history of campaign responses could give an optimization system more material to learn from than a workflow that considers only placement or audience categories.

Tatari says the system learns from campaign outcomes rather than relying only on where ads ran. That’s the core of its argument: exposure data can show where an ad appeared, while outcome data can help assess what followed. Connecting the two is difficult across TV formats, where measurement may involve different publishers, viewing environments, and conversion paths. If you’re working in this space, the important question is how well those signals can be compared and attributed across campaigns. The data is aggregated and anonymized, but the company hasn’t supplied details here on how it handles differences in measurement or how it tests the model’s predictions.

That distinction matters because optimization on direct-sold inventory depends on connecting exposure to results across TV formats. Tatari’s bet is that its measurement history and direct publisher integrations can make impression selection more responsive than conventional buying workflows. The approach also rests on a feedback loop: campaign outcomes inform later decisions, and later decisions create new outcomes for the system to assess. That can be useful, but the loop is only as informative as the measurements entering it. A system may learn to favor patterns associated with past success without proving that the same patterns will hold for every advertiser or campaign.

Early campaign results

Two early adopters have reported gains from campaigns using Tatari Max. Rylo, an AI communications platform serving Deaf and hard-of-hearing people, saw conversions rise 36 percent. Performance activewear brand Rhoback reported 32 percent more purchases. Those results are a promising first indication, but the source doesn’t specify the comparison period, campaign size, or whether the figures reflect a controlled comparison. Without that context, the results show what the companies reported, not a general performance guarantee.

“A lot of what the industry calls AI right now is basic rules-based automation; Tatari Max is true AI intelligence: it learns from measured outcomes and gets smarter with every campaign,” said Philip Inghelbrecht, Tatari’s CEO and co-founder. He said the company had spent years building its measurement and data infrastructure and integrating with publisher ad servers. The early results offer a first signal of the approach’s potential; the larger test will be whether it can deliver consistent gains across campaigns and the full TV mix. The distinction Inghelbrecht draws is also a claim about how the product works, and buyers will need evidence about performance over time to assess it.

What the approach could mean for TV buying

Outcome-based selection could make direct TV campaigns more adaptive, especially if an advertiser can use measured results to guide where later impressions go. That would narrow a familiar gap between digital advertising, where campaigns often adjust rapidly, and TV buying, where plans can be shaped by negotiated inventory and longer execution cycles. Here’s the thing: a faster decision process is valuable only if the measurement behind it is reliable and the available inventory gives the system room to act.

This is more significant than it looks if Tatari can show that the system works across different advertisers, objectives, and TV formats. For now, the evidence presented is a set of company data claims and results from two early adopters. Buyers will want to understand how gains are measured, how performance compares with a campaign’s prior approach, and whether results persist beyond initial deployments. The test isn’t just whether an algorithm can select impressions in real time; it’s whether that selection creates repeatable value in the conditions advertisers actually face.

And yet, the model has a clear appeal: it brings a performance focus to inventory that has often been bought through direct relationships and planned placements. Tatari’s publisher integrations and measurement history may give it a foundation for doing that across a wider share of TV. The numbers here are encouraging, but they don’t settle the question. More evidence across campaigns will show whether Tatari Max is a practical change to buying workflows or an effective system in a narrower set of cases. A promising opening, with the proof still to come.