Halo effect

The canonical chain: broad-reach media (TV, video, audio, out-of-home) creates demand and awareness; that demand expresses itself later as branded search queries, direct visits, and higher conversion rates on performance channels. Last-click measurement assigns all of it to the last touch, which is almost always a bottom-funnel channel. The brand channel that caused the demand looks expensive; the channel that harvested it looks brilliant.

Modern MMMs attack this in two ways. Some model cross-channel paths explicitly, letting TV spend enter the branded-search equation as an input.¹ Others rely on the decomposition mechanics: if TV's adstocked spend co-varies with revenue beyond what search spend explains, TV gets its credit directly.² Either way, the model needs variation to see the effect; if TV has run at a constant level for two years, its halo is statistically invisible and lands in base.

What happens if halo is ignored: the reallocation spiral. Measured naively, brand looks weak, so budget moves to performance. Branded demand then thins, performance channels get more expensive per conversion, total efficiency drops, and the response is usually another round of the same cut. Brands discover the bottom of this spiral empirically, about two quarters after the CFO stops asking questions.