What eBay learned when it switched off its brand ads

Most of the money spent on brand keywords buys clicks you were going to get anyway. eBay checked, and then published the result.

Written by the Colleviate Strategy desk

What did eBay switch off?

Three researchers, Thomas Blake, Chris Nosko and Steven Tadelis, asked eBay a plain question: what happens to sales if the search ads stop? They answered it by stopping them. The paper, Consumer Heterogeneity and Paid Search Effectiveness, appeared in Econometrica in January 2015.

The first result is the one worth remembering. When eBay stopped paying for ads on searches that contained its own name, the sales didn't go away. People clicked the free listing instead. The authors report no measurable short-term benefit from the brand ads.

Put plainly, the ads were being credited with sales that would have happened anyway. The ad platform's report would have said otherwise, because a report counts a purchase that follows a click as a purchase the ad made. That is the gap the ROAS reality check describes, shown at the scale of one of the largest advertisers online.

What about the other keywords?

The team also paused bidding on non-brand keywords, the generic searches, for 60 days across a 30 percent sample of eBay's US traffic. They report that the measured returns were a fraction of what a standard, non-experimental estimate suggested. The reason is plain once you see it: people who search for a product are, more often than not, people who were already inclined to buy it. One version of the abstract adds that new and infrequent users were positively influenced by the ads. Read that as a clue about where ads do work, not as a rule.

What does it mean for you?

Concede the obvious first. eBay is one of the best known names on the internet, and a brand nobody has heard of will see something different. The authors say as much: ads reach people who don't know you yet. So this isn't a case for turning your ads off. It's a case for knowing which of them are working.

The useful split is the one in branded and category search. Brand searches are people who already know you. Category searches are people who don't. They deserve different budgets and different tests.

Run the test yourself

1. Pick two cities that behave alike, with similar sales and similar search volume for your brand.
2. Pause brand ads in one of them for four to six weeks and leave the other alone.
3. Judge by total conversions from brand searches, paid and free together. The platform's own number will always say the ads worked.
4. Check who else bids on your name. If a rival buys it, their ad may sit above your free listing, and pausing could cost you sales. That's our reasoning, not a finding of the paper, which is why you test it.
5. Move the saving to something that reaches people who don't know you yet, and measure that the same way. The mechanics of the account are in Google Ads account structure, and the same logic applies to retargeting.

The paper's value isn't the result. It's the habit: switch something off, in a small place, for a short time, and look at what really happens.