“Traffick: To make this as clear as possible, I’m trying to get a sense of whether a number of the “other relevancy factors” you look at for AdWords actually affect ad rank, or whether you’re still letting CTR history and minimum bid decide together where ads are ranked on the page, with the other factors coming into play only with regard to setting minimum bids?
NF: This is where it gets complicated. Actually, there are two quality scores. The first is for setting minimum bids. It’s less accurate. [Traffick: Thus, we assume, predictive where limited data is available on CTR History.] It takes into account landing page quality. The information we use for ranking ads is much more specific.
Traffick followup: So, I’ll assume a whole range of factors are involved in
deciding ad rank but that along with bid, CTR History is, as Google has said in the past, “predominant.” But you’re saying that landing page quality does not currently affect ad rank; landing page quality only affects minimum bids.
NF: We’ve been debating this internally, and it’s my feeling that our users will benefit even if we push the landing page quality measure into the quality score that affects ad rank. Basically, I advocate all of the factors being used in the actual rankings of the ads, not just affecting minimum bids. So I’m pushing for landing page quality being included in the ranking formula and I think you’ll see that soon.
Traffick followup: What are some of these other factors besides CTR?
NF: While much of this is proprietary, I can say that many of them are just different cuts at predicting CTR.
Traffick: How can you provide more information to advertisers who have low quality scores, and is there some way advertisers can appeal high minimum bids that might be based on “algorithmic false positives”?
NF: Google is currently exploring ways of providing more of this type of information to advertisers. The problem is, if we give too much information about the process to the bad guys, they’ll turn around and use that to circumvent the process.
Traffick followup: So they try to “fix” it, but the intent is not to really fix it.
NF: Right, the intent is not to really fix it. In any case, Google is running a test now — we haven’t formally announced it yet, but I guess I’m going to tell you now — showing in the front end, quality score information to the advertiser. This is sensitive. Potentially it helps the bad guys optimize, so we have to be careful how we implement this. As for the editorial process and appeals, today there is an escalation process, and Google can correct mistakes if it sees that a mistake has been made. There are actually few enough of these appeals that I get to see all of them personally.
Bonus question: Do you use things like account history (as in spend, or length of time with an account) in the quality score algorithm?
NF: No. Google doesn’t believe in setting up perverse incentives when it comes to improving the user experience. So we don’t use total spend or time as incentives.”