In possibly the most significant announced change to AdWords, the Quality Score is changing. This should, within a month or so, cause improvements in conversion if the details are correct. It would be quite possible to mess this up. I suspect that there will be some weeks of agony as first, users will assume that this starts now - it looks like it will take weeks to roll out; and also Google seems to reset the learning for the AI when they make these big switches. So the system will probably have to learn again, what matches with what - while maintaining some kind of account history.
I’m expecting a busy September, troubleshooting.
I’ll be thinking about this for the next while… results when I get some!
Speculation…
Hmm. reading between the lines, this is probably for optimising medium to larger accounts (more than tens of clicks per day). Why?
Wild guesses, since I don’t work for Google… I’ll bet that they are using some kind of multivariate real time data analysis technique. Possibly something like Stochastic Perturbation with Simultaneous Annealing (SPSA) or the Taguchi Method. I’ve certainly had problems using these techniques on low volume accounts. When you get down to a few tens of impressions, it takes a long time to establish any statistical basis for action.
Making optimisation more complex, AdWords operates in a number of different regimes. An advert may be shown on Google’s properties before editorial review, and on the Google and Content Network after review. That substantially modfies the performance. Google knows exactly when an advert is reviewed… But advertisers have to guess based on seeing impression volume changes and clicks from non-Google properties. That means you really want to tie either a third party redirector, a JS Web Beacon/Page Bug to for landing pages, or hook into the web server log files, just to infer when the review has been completed… If you don’t, you get skewed results.
Small bid changes can also affect the regime - so bids can affect not only position, but also whicj searches show the advert. And other keywords interact, too. Drop the bid on one broad matched keyword, and you may find hat impressions ramp up on another broad matched keyword, for the same set of search queries.
I can imagine that Google *could* do a better job of optimising, using data that they uniquely hold, than any kind of external analysis. I expect that their main method of optimisation will be to optimise their inventory rather than your ROI, though. Yeah. I’m sceptical. Still.
