adCenter Quality Score Observations
We have examined historical campaign data as a first step in the effort to better understand the dynamics of adCenter’s version of Quality Score. To recap, the open questions we focused on:
- How does (position normalized) CTR affect the CPC that you pay in adCenter?
- How much influence does quality of landing page have on CPC (and position)
We reviewed performance data over the course of a 6 week period from Jan-Feb 2011. The Ad Group observed has KW destination URLs so we were not able to extract any signals regarding ad landing page influence on CPC & position in this setting. As a bit of additional context, the makeup of the campaign we observed is as follows:
* 3 unique creatives running against 150 KWs in the same ad group
*The campaign had been live for several months and received considerable traffic prior to the time period reviewed
*For the period reviewed each ad received several hundred thousand impressions
For the period reviewed, key data points were as follows:
| Ad 1 | Ad 2 | Ad 3 | |
| CTR | 5.56% | 1.61% | 1.44% |
| Avg CPC | $0.14 | $0.13 | $0.14 |
| Avg Pos | 3.12 | 3.19 | 3.51 |
| Impression % | 33% | 37% | 30% |
Observations:
We have certainly been intrigued by the results. While it appears there is some correlation between CTR and avg position, we are surprised a delta of the observed magnitude – against considerable volume – did not result in a bigger delta in avg position. Even more surprising is the near flat avg CPC for each creative and the division of impressions within the ad group. Based on what we have observed here, it does not appear adCenter values CTR in the same way AdWords does when awarding CPC discounts.
Even if adCenter is not discounting CPC based on CTR there is still clear value in conducting ad testing and optimizing campaigns toward creatives with high CTR. Logically a user is more likely to click on an ad with a higher CTR than a lower CTR. Therefore, the more impressions that are sent to higher CTR ads, the more volume the ad group should accrue. That said, if the true campaign success measurement is ROI, CPA or another metric influenced by a conversion, you will obviously want to take any relevant metrics into account as well when making your ad optimization decisions.
It is important to call out that the KW profile of the ad group may have colored the result we saw to some degree. Specifically, ad group contains high & low volume KWs and both brand & non-brand keywords. Obviously we do not have data detailing performance of specific KW-Ad pairings to understand these dynamics have controlled for these factors in our follow up testing .
Our more controlled test referenced here is currently running but not accruing as much data as we would like to make an definitive call on results. Stay tuned as we work through some ad serving issues with AdCenter.
In summary, to this point in our investigation our 2 primary conclusions are as follows:
* Large ad CTR deltas do not necessarily result in corresponding CPC & Avg Pos deltas
* Advertisers should be cautious about relying on adCenter for adwords type ad optimization (against CTR) at this point
Stay tuned, we are looking forward to analyzing results of our follow up testing and will be sharing our observations.
Interesting. Did the ads all run at the same time? I thought adCenter switched it’s ad rotation from rotate evenly to optimize for best performing ad. I’m just surprised your impression share was pretty even across each ad.
Thanks for the note. Ads did run at the same time — from what we understand ad rotation does exist on AdCenter but even so, the capability only exists on the O&O (yahoo.com and bing.com) so most traffic will still come from equal rotation. We were a little surprised that the rotation was basically even, given the big performance deltas, but points to a big opportunity for AdCenter to improve the quality of its marketplaces by enabling optimization across all searches.