Interviews, insight & analysis on digital media & marketing

Football had a new ad break this summer. Did anyone prove it worked?

By Jason Hicks, GM, Kochava for Advertisers

At the 22nd minute of each half in every 2026 World Cup match, the referee blew the whistle for a scheduled hydration break. It was FIFA’s first structured mid-match commercial window in tournament history, written into all 104 games. Fox cut to ads. Streaming placements in that slot commanded between $60 and $120 per thousand impressions, with the break sitting near the top of the range.

On paper, it looked like a clean opportunity. A live audience, globally engaged, phones already in hand, and two minutes of premium inventory that hadn’t existed four years earlier. The harder question, one the industry is now sitting with, is whether anyone could connect that moment to what a viewer did next.

The broadcast picture complicated things from the start. Matches ran simultaneously across Fox, Telemundo, Peacock, Tubi and YouTube TV, each generating separate data streams. Fox and Telemundo combined were projected to pull in $850 million in World Cup ad revenue, yet 41% of buy-side leaders said CTV is underrepresented in their measurement models. In a tournament where streaming was the primary premium surface, that was a significant gap to carry into a $10.5 billion ad quarter.

The hydration break was only part of the problem. Think about what a typical fan’s tournament visit looked like. They would arrive at LAX to a branded countdown clock in the arrivals hall. The next morning there was a streaming pre-roll. On match day, a digital billboard on the light rail to SoFi Stadium, a geofenced notification at the fan zone entrance, and then a $120 shirt paid for in cash at the merchandise counter. Five surfaces across four days, and not one platform’s dashboard could reconstruct that path from start to finish.

The physical surfaces were where most measurement stacks broke first. EMARKETER put total US out-of-home spend at $11.3 billion this year, driven in part by World Cup activation. Nielsen’s data from last summer’s Gold Cup final found over 30% of the audience watching from bars, fan zones and public spaces, generating no device ID, no login, no addressable signal. They were in the room. They saw the ads. Most attribution tools had no record of them.

The commercial stakes made this more than a technical problem. Numerator’s research put purchase intent among World Cup viewers at 89%, higher than both the Super Bowl and the Winter Olympics, with projected consumer spending of $7.5 billion across the 39-day window. Budgets at that level cannot be reconciled across four separate dashboards and called good enough.

Marketing mix modelling came back into focus precisely because of problems like this. EMARKETER reported that 47% of marketers planned to invest more in MMM this year, because it can account for channels that leave no digital trace (out-of-home, broadcast, in-venue), and build a picture of campaign performance that outlasts the event itself. It provides the wide-angle view that individual platform reporting never could.

What MMM does not do on its own is prove causation. That’s where incrementality testing earns its place. The question it answers is not what a campaign added up to across six weeks, but whether the sequence of impressions across those five surfaces changed what the fan did, or whether she was already heading to the merchandise counter. Running both together is the difference between knowing what your spend achieved and estimating it after the fact.

The trophy was lifted on 19 July. The questions that followed were predictable: which surfaces drove the conversion, whether the hydration break delivered the uplift its price tag implied, and what the out-of-home exposure was worth when the stadiums emptied. Brands with the measurement infrastructure to answer those questions clearly will carry real knowledge into the next major event. The ones without it are left with an expensive approximation of what they spent six weeks buying.