Adwords Ads Split Testing Paradox
December 13, 2007
It sound very complicated, but its rather quite simple…
Split Testing Procedure, Simplified
- place 2 or more ads per adgroup, I prefer two, as in this example
- check history for a good amount of traffic
- edit the poorer performing ad, in this case the second one
- that’s it, really.
As you can see in this example, a very minor difference in the ad text had a significant effect on CTR, 7.25% versus .75%
Wow wa weewa
If you are not performing this A/B Ad Split Testing, you are certainly missing out on a very fundamental optimizing technique.
Here is the screenshot after I edited the second ad, as see, the second ad has had its ad history wiped clean:
Adwords A/B Split Test Paradox
In order for the ads to compete equally, each ad should be ranking at the same position on the search results page, but the newly edited ad has no history, and will therefore rank lower, creating the paradox.
There is no solution to this paradox, except waiting enough time before comparing results, to give the new ad enough history to rank evenly with the old ad. How long is enough? Probably 2 days minimum, assuming there is some traffic, but its not clear. Generally I don’t optimize ads more than 2x/week, so the paradox is less relevant; just be aware of the issue and don’t compare too soon.





0 comments:
Post a Comment