“The concept behind it is that through incrementality testing you are able to calculate sort of a score between what you actually see as direct tracked data and what was more hidden — maybe you over-attribute to search and under-attribute to social ads that come earlier in the funnel. You use that score to modify your statistic model.”

Media Mix Modeling Needs Incrementality Calibration to Be Actionable

Raw media mix models have a known flaw: they assume a static baseline and can't fully account for cross-channel influence. Deezer's approach is to use incrementality tests — geolift, conversion lift, brand lift — to calculate adjustment scores for each channel, then apply those scores to the MMM. The result is a calibrated model that better reflects the true contribution of upper-funnel spend on final conversion events.