They switched paid search off and sales barely moved
What was done
eBay ran a large controlled experiment on their own paid search, published by Blake, Nosko and Tadelis in Econometrica in 2015. Branded keyword ads were switched off in test markets while matched markets kept running.
What was done well
The design itself. They used real markets, matched controls and their own revenue rather than a platform dashboard. Almost nobody does this, which is why the paper is still the reference twelve years later.
What went wrong
Before the experiment, the account was judged on attributed return. The naive measurement said the spend returned 4,173%. It had been buying clicks from people who were arriving anyway.
What should have happened
Run the switch-off on branded terms first, in a subset of regions, for a defined period. Branded is where intent is highest and therefore where attribution flatters most.
What the right plan does to the economics
On the same spend, attribution said plus 4,173% and the experiment said minus 63%. For an account spending $40,000 a month on brand terms, a result like that turns roughly $480,000 of annual spend from a reported profit centre into a cost with no measured return. The money does not disappear: it moves back into channels that create demand rather than harvest it.