A bidder enters an auction for the query "running shoe" and bids the bare minimum because it can't see that the user is near a retail establishment. What should the marketer do to improve the bidder’s conviction?

The marketer should use store sales and app history signals as instructions. The bidder is not malfunctioning: it is bidding conservatively because nothing in its data says this user is valuable. Someone standing near a shop who might walk in and buy looks identical to any other searcher if the account only measures website purchases. Feeding in store sales data, store visit conversions and app activity gives the model the missing evidence, so it can recognise the pattern of a nearby, high-intent shopper and bid with conviction. This is the human-in-the-loop principle in its most practical form. When automation behaves timidly, the fix is almost always to supply better instructions and richer conversion data, not to override the bidding or throw budget at the problem.