Basically, maximize ROI with real data. In ecommerce situations you may be able to report sales data to G, but what that is in the moment of the transaction and what that ends up being after shipping complications or cancellations can be very far apart.
For instance, I manage about 300 thousand keywords and match types in Google and Bing. I have a set of rules that looks at real data and makes adjustments -- then I manually look at the changes to make sure they make scene with what I know about whats going on in the market, seasonality ans such.
Google offers automatic bidding where you can set up cost per acquisition, but the best you are going to do with that is approximations based on the data you are able to feed Google. You will have to rely on their bidding algorithm though and most of the independent data I've seen out there shows that it doesn't do as well as manual bidding.