Tatari Max uses AI to pick which TV ads each household sees

Drawing on demographics, shopping habits and device IDs, the engine rates each impression, one showing of an ad, with what Tatari calls an affinity score: its estimate, from past viewers’ responses, of how well that slot suits a given brand, AdExchanger reports. Tatari says the engine learns from more than $9 billion of client spend it has managed, 180 million conversion events it processes daily and the measured results of over 23,000 ads. In Tatari’s trials across dozens of campaigns, the median campaign improved its incremental cost per acquisition, what each extra sale caused by the ads costs, by 19 per cent, the company reports. Activewear brand Rhoback saw 32 per cent more purchases over a one-month run of one Hulu ad, and hopes the engine will learn which of its ad versions suit which viewers.

Making those versions has become cheap. Our 29 October price check puts a Veo 3.1 Fast clip of four to eight seconds at $0.40 in Google AI Pro’s monthly credits; Kling 3.0 at 1080p with audio comes to about $0.91 for five seconds for a member buying top-ups, on our 23 September check.

With a short Veo Fast clip at $0.40, ten alternative openings for one TV spot cost about $4 to generate on our sheet, so the generation bill for the several versions Rhoback hopes Tatari Max will match to viewers is a few dollars, before editing and broadcast checks.