Every automated bid strategy in Google Ads is an optimisation engine pointed at a target you define. That is the whole deal. You name a conversion, you attach a value, and the system spends your budget finding more of it.

Which means the most consequential decision in the account is not a bid, a budget or a keyword. It is the definition of the conversion — and it is usually made casually, early, by whoever set up the tag.

What a loose definition costs

Suppose “conversion” currently includes form opens, phone-number clicks and three-page sessions. All three correlate loosely with interest. None of them correlate with revenue.

Smart Bidding does not know that. It finds the audience most likely to open a form and never submit it, and it gets very good at buying them. The account reports a falling cost per conversion. The sales team reports nothing changing. Both are telling the truth.

The three questions worth answering in writing

Before scaling anything, answer these — on paper, with the person who owns the revenue number:

  1. Which single action, if it happened more often, would make the business more money? Not which action is easiest to track. Which one matters.
  2. What is it worth? An average is fine. A rough average beats no value at all, because without one, every conversion is implicitly worth the same.
  3. How will we know it happened? Including the cases where it happens three weeks later, in a CRM, after a phone call.

When the sale happens offline

Most B2B and high-consideration accounts have a gap between the lead and the revenue. Offline conversion import closes it: lead IDs go into the CRM, and the outcomes come back into Google Ads with real values attached.

It takes a week of setup. It changes what the bidding algorithm is optimising toward from “someone filled in a form” to “someone became a customer worth €4,200.” Nothing else you can do to an account has that leverage.

A reasonable sequence

Fix the definition. Attach values. Verify the data arrives and is not double-counted. Let the bid strategy relearn for a full cycle. Only then increase budget.

Scaling before that step is just buying more of whatever you accidentally asked for.