Optimizing your Google Shopping product titles based on search term data is a tricky proposition. The AdWords Search Query Report for Shopping Campaigns gives you impressions, clicks, conversions, and their corresponding click-through-rates (CTR), and conversion rates (CR).
Using CTR or CR is flawed because items with only a few clicks tend to have artificially high CTR’s. For example, a 25% CTR doesn’t mean much if there are only 4 impressions per month.
Using clicks or conversions is flawed because they tend to express what’s already working rather than uncovering opportunities for improvement.
We need a better metric to find where the real opportunities are, and judge the real expected performance; one that scales from a few impressions to thousands.
I think we can calculate an expected level of performance (blue line in below image) for each search term based solely on its number of impressions. Then we can calculate how much its real clicks and conversions are above or below that expected level. A search term that has a 22% CTR and 4 impressions may be below this curve given its long tail specificity. But another search term that has a 19% CTR and 50 impressions may be doing great, and therefore would be a candidate to include in more product titles.
Conversion rate seems to follow a power curve relative to CTR which again provides a relative comparison to expectation.
Or we can relate conversions straight back to total impressions.
Once we have calculated our expected SQR performance curve, we can analyze each term or each Google Shopping campaign and quickly deduce where our top performers are, where up-and-coming products and terms are, and problem campaigns or products that are wasting money for less return.
Due to the statistical noise on CTR and CR for low frequency impressions, a possible refinement would be to calculate the final above/below performance difference based on a volume-adjusted standard deviation from expected curve instead of a straight percentage.