How negative search remarketing does (not) work

How negative search remarketing does (not) work

Some days ago, tickled by a very good post by Rossella Cenini (a top contributor of the official Italian AdWords Community) I discovered to my great surprise that it is not so easy to exclude from your search campaigns one or more user groups tracked among your remarketing lists, even if that segment is explicitly excluded.
Negative search remarketing in AdWords is quite tricky to configure, although it seems clear how to do it. Before telling you the workaround cleverly suggested by Rossella, let’s step a little back to understand how search remarketing works (both in positive or negative targeting).
Remarketing in the search network allows you to push (or avoid showing) ads related to specified queries only to certain user groups tracked by AdWords native remarketing tag.  There are many cases in which this can be useful. I personally have identified at least a couple of them, both for positive targeting and negative one.
When acquisition costs of new customers are high and repeated buying is frequent it may be convenient to push ads to returning customers even when they are searching for queries more generic than the ones we normally focus on. For instance, if you are running campaigns for a booking portal it could make sense to address ads to searches like “hotel offers” or “accommodations in Italy” (and similar) when performed by returning customers (the same queries targeted to undistinguished searchers would be in the majority of cases a real quality score and CPCs hara-kiri).
As for negative search remarketing, more than once I found myself thinking to exclude from showing search ads for a while to users who have already converted. If one has already bought my product and goes back to Google looking for the same thing is very unlikely to buy it again, unless you sell repeated purchase goods or services, such as shoes or bags (or you are Zalando! ;-).
And here I was wrong! On one hand because search remarketing exclusions do not work well by now, and, on the other hand, because a lot of repeated visitors converted extremely well. To understand this I looked into an Analytics custom segment with converted users coming from that search campaign and compared data before and after the exclusion of the list.
search-remarketing1
If exclusion worked well there should be a lot less returning visitors in the period 5/21-31 (when more than 2.000 converted users where, in theory, excluded from search ads). But no sensible difference is evidenced before and after the exclusion.
Even the analysis of acquisition channels evidenced that nothing has really changed after blocking the converted list of users. The vast majority of users is still coming from paid search clicks and all other sessions sources, including referrals and direct URL calls, have very small statistical significance.
search-remarketing2[clear]
In this case the missing exclusion, which I mistakenly thought would have lowered conversion rates, actually turned out to be a positive thing, because these obsessive-compulsive searchers have very high conversion rates!
Web users probably tend to forget quite easily brand names, even when they bought something on your site, and constantly search for your services in the engines (in this campaign brand names were specifically excluded with negative keywords!).  If your ads are not there, there is no guarantee that they will find the way to your site again!
Tests carried out both by Rossella Cenini and Luigi Sciolti (another very good AdWords Italian professional) confirmed that if the list is moved from exclusions to search remarketing positive lists, with a -90% bid adjustment, numbers of returning visitors are greatly reduced. So if, at the contrary of mine, it makes sense for your campaigns to prevent a group of your visitors to see your ads, that is the right way to do it (at least by now).
The final consideration is that in web marketing you should never take anything for granted and always keep your eyes open to confirm your initial hypothesis with data and other reliable colleagues experiences (thanks very much to Rossella Cenini, Francesco Tinti, Filippo Trocca and Luigi Sciolti 😉
 

Gianpaolo Lorusso
info@gplorusso.it
No Comments

Post A Comment