eBay has been one of the largest advertisers on Google, but that might not be the case for much longer. It decided to do an A-B split test to determine how many of those clicks they would have eventually seen even without those paid placements; even going so far as to go dark in 30 market areas to provide a control. In a study conducted with eBay Labs along including fancy degree holders from Berkeley and U. Chicago, it showed that it only made back about 25 cents on the dollar spent.
The study shows that brand ads - and by that they mean ads that focus on the brand names of the products to be theoretically purchased, rather than "branding" ads - can be efficiently effective for potential new users to the retailer, but tend to be unnecessary for those who are already familiar with the retailer. eBay, more than most, would suffer from a high familiarity ratio, thus making its relative efficiency low.
One thing the full version of the study appears to miss, however, is that very high clickthrough rate experienced by an ad due to a specific brand reference that is common to the search term may have another financial benefit to the advertiser: increasing the "quality score" of the ad campaign, and thus reducing the expense of other clicks in the campaign.
eBay bids on a universe of more than 170 million keywords. It spends more than $50 million a year on online advertising. The study's conclusion indicates that a simplistic measure of return on investment would show it made back five times its media costs, but that wouldn't take into account the fact that most of those buyers would have bought on eBay anyway. Factoring out that, plus some other cyclical noise, the study writers found a negative ROI of 75 percent.
The study found that customers who had not yet fully embraced eBay as an acquisition platform (making six or fewer transactions in the past 12 months) showed a positive sales response to the advertising. People with seven or more transactions showed no significant change.
Using another metric, they found that there would be a positive effect for those users who had not been to the site in the past four months, but no effect for those who had visited more recently. That last metric is interesting in that it could be associated to a retargeting campaign explicitly designed to hit users in the proper visiting timeframe.
On factor making that incremental business less attractive is the notion that these influenceable people are eBay's worst customers, and likely have the least value per user, making the predicted efficiency of advertising even less. The study graphs the number of transactions purchased by paid search customers, and it is all bunched to the low side of the graph.
That attribution is a difficult metric to suss out (the determination of which ad communication was responsible for a particular purchase) is not knew to the online world, or the traditional media world for that matter, but the study may raise eyebrows as to just how much the online practitioners may be mimicking the traditional media players in conceitedly over-estimating their own influence.