Myth: pausing brand search is reckless — those conversions would be lost

Myth: pausing brand search is reckless — those conversions would be lost

The question: brand-keyword campaigns convert at a spectacular cost-per-acquisition, so any attribution model crowns them. Does that mean the conversions they're credited with are actually caused by the ad?

What the experiments found: this is one of the few attribution questions with a famous, replicated causal answer. eBay's large-scale paid-search experiments, and subsequent brand-bidding holdouts across many advertisers, found that a large fraction — frequently most — of brand-search conversions are not incremental. When the brand ad is paused, users who searched the brand simply click the organic listing directly below and convert anyway. The paid click was cannibalizing a free organic click.

The nuance: this is the purest correlation-causation trap in the discipline. Brand search sits at the very bottom of the funnel, capturing demand that already exists, so every observational model — last-click, DDA, Shapley, Markov — assigns it enormous credit. None of them can see that the conversion would have happened for free. Incrementality is not zero in every case (competitor conquesting on your brand term, weak organic presence, and ad-vs-organic real-estate effects create real defensive value), but the default assumption that brand-search credit equals brand-search value is wrong.

Bottom line for practitioners: never trust an attribution model on brand search. Run a brand-bidding holdout (pause in matched geos, watch total — paid plus organic — branded conversions). Only the incremental delta, not the attributed CPA, tells you whether the spend is defending revenue or buying clicks you'd have gotten free.