Should Google Ads optimise for purchases or sign-ups?

The standard advice, when a campaign bidding for purchases looks slow, is to add something that happens more often: a sign-up, a trial, a lead form. The argument is that the bidder learns from conversions, and more conversions means faster learning.

The platforms do tie their rules to volume, so the advice is not folklore. Google's Target ROAS page sets a floor for Search and Shopping campaigns of at least 15 conversions in the past 30 days at the conversion tracking level. Google's learning period page says it can take up to around 50 conversion events or 3 conversion cycles for the bid strategy to calibrate to the new objective, although it can be faster depending on the amount of conversion data present. Meta, among its fixes for a learning limited ad set, suggests choosing an optimisation event that occurs more frequently, for example moving from purchases to add to basket.

In Google Ads, the Conversions column shows the conversions you've received across your primary conversion actions, and Google says the data in that column is what your bidding strategies use. Google's Target ROAS page adds that Smart Bidding will learn across all conversion actions reported in the Conversions column, even if they're configured for different goals. We read that as meaning a sign-up included in the column is something the bidder is now trying to get, not only something you are now counting.

Google's own conversion tracking page is more cautious than the advice. In its guidance on the conversion count setting, it says to ensure that only bottom-of-funnel actions are set as primary, to avoid inflating counts.

The next step is arithmetic, not anything Google publishes. Cost/conv. is your total cost divided by the number in the Conversions column, so once sign-ups are in that column the cost per conversion reads lower than it would on sales alone, whatever happened to sales. Read that figure after the change and the goal looks like it worked before anyone has checked whether more people bought.

The figure that answers the question is cost per sale, and it splits exactly into two halves: cost per sale equals cost per click divided by the share of clicks that become a sale, provided all three come from the same spend, clicks and sales. This is arithmetic, and it is how we read every goal change.

A goal change can move either half. If cost per sale fell because the click-to-sale rate rose, the new goal found people more likely to buy. If it fell because clicks got cheaper while the click-to-sale rate stayed flat or fell, the new goal bought cheaper traffic, and the saving lasts only as long as those clicks stay cheap. The two need different decisions, and the headline cost per sale cannot tell them apart.

This is our own observation, not either platform's, and we do not put a number on it because it rests on too few accounts for a number to mean anything. Where we have added or switched to a shallower goal next to the sale and then run that split, the cost per final sale did fall. The gain came from the price of a click. The share of clicks that went on to buy stayed flat or fell.

In the same accounts, the sale-only goal had run on its own for months before the change without the strategy failing. Lack of signal was the reason given for the change, and the history did not show the strategy starved of it.

This is our practice, not Google's or Meta's. Before adding a shallower goal, we check whether the sale-only setup is actually failing, against the platform's own floors and statuses, rather than assuming it is. After the change, we report cost per sale, never cost per conversion, and we split it into click price and click-to-sale rate before calling the change a win.

If the gain is in the rate, keep the goal. If it is all in the click price, the shallower goal may still be worth keeping, but as a decision to buy cheaper traffic made knowingly, not as proof that the bidder needed more signal.

On value bidding the question changes shape. Target ROAS weighs conversions by the value they report, so, on our reading, a sign-up given a token value moves the conversion count far more than it moves what the strategy is chasing. Our Target ROAS entry explains why we judge those campaigns on value per cost instead.