Impressions Strongarm Relevance in adCenter Quality Score

A couple of days ago, Jessica’s latest episode of Ticker Tuesday focused on an issue that’s of pretty substantial interest to search marketers of late: adCenter quality score.

Since Yahoo and MSN became one, many paid search advertisers have noticed substantial differences in their performance that are difficult to account for. Given the amount of time we’ve had to adjust our expectations (adCenter handles keyword match typing differently than Yahoo, and keyword duplication differently than Google, etc) and strategy, it can be complicated to understand why some campaigns, ad groups, and keywords perform beautifully in adCenter while others seemingly can’t gain an impression. I think everyone’s hope was that the visibility of adCenter’s quality score determination would help us to find opportunities to modify our accounts to match MSN’s expectations, which are clearly not the same as Google’s.

Jessica did a great job on Tuesday explaining the basics of adCenter’s quality score determination process, but I wanted to follow up with a little more discussion of some of the insights we gained from the quality score webinar MSN recently held.

One of the most interesting, and confusing, parts of the webinar was the discussion of account hierarchy and how quality score is determined relative to it. Basically, they broke down that in terms of influence on quality score,

High impression keywords have more influence than low within their ad group.

High impression ad groups have more influence than low within their campaign.

And high impression campaigns have more influence than low within their account.

The takeaway for us is you can only view quality score at the keyword level, but ensuring that a keyword has a good quality score in and of itself isn’t really adequate and you’re going to have to get into some math that’s more complex if you want to leverage quality score visibility to improve your account’s performance.

Why? High-impression, high quality score keywords have a lot of influence on the average quality score of your ad group overall. That means that low quality score, high-impression keywords also have a lot of influence on the average quality score of your ad group, but you can’t always turn them off because we all know that sometimes those are still high performers for one reason or another. You should try to improve their quality scores as you can, but MSN goes a step further and says that because lower-impression high quality score keywords are negatively impacted by (“susceptible to”) the restrictions caused by being grouped in an ad group with the high-volume, low-quality keywords, you should take them out and group them separately.

Furthermore, an ad group can have all-10 keyword quality scores, but if its campaign is full of higher volume ad groups with awful keyword quality scores (or just a few very high impression keywords with awful keyword quality scores), the keywords with quality scores of 10 might still not gain the traffic they otherwise could as a result of their bad neighbors in other ad groups. To determine the relative influence of a keyword, you have to therefore consider its quality score, its impression volume, and how both of those things compare to the same metrics for the keywords around it.

The simplest application of this information is to identify any low-quality score, high-impression keywords in your account and actually review their value to you. Possibly try pausing or moving them to test whether the performance of the other keywords in the ad group/campaign/account can be improved by their removal (and thus the removal of their large influence on all levels of quality score). MSN did mention their time frame for considering quality score determinants (recent data is given more weight than historical, but adCenter does look back up to 3 months and if volume varies substantially, will weigh recent data less) so keep this in mind if testing keyword removal or restructuring.

A more complicated application is to manually calculate the relative impressions (and therefore relative weight) and quality score averages at each account level and try to structure your account to take advantage of this weighting process. I understand this and I will probably look into it, but a part of me feels like what does this have to do with relevance and improving user experience? Grouping keywords by impression volume rather than relevance to one another and my products seems like a convoluted workaround to achieve high quality scores. If MSN’s quality score formula requires action that has nothing to do with increasing relevance to get your ads to show and lower bids, then there’s a problem with that formula.

I am personally having quite an interesting time in a couple of my larger accounts in trying to restore volume in MSN to pre-merger levels, and have not had exceptional success as of yet. Based on the data presented in the MSN webinar I am considering launching a very different strategy in which I identify keywords that have traditionally driven performance in MSN and substantially reduce the active keyword volume in this account to allow for more straightforward visibility of how each higher-volume keyword is performing, and slowly reintroducing the other keywords to see how they influence one another and test the impact that separating based on impression volume can have. Impression volume has been so low recently anyhow that I am not sure that the expansive keyword strategy imported from Google is doing me any good anyhow. Any little bit of information that gives some clues as to why adCenter behaves as it does is helpful, so I’d like to test a bit here. We’ll see.

If you would like to see MSN’s quality score webinar, it is interesting though long (about 60 minutes). It created more questions than it answered for me, but that’s not always a terrible thing. If you attended it, or watch it now and have other insights let us know!