How Target ROAS treats a low-value conversion goal, and what removing it really costs

We met this on one account, so this post carries no figures from it. A campaign on Target ROAS was bidding on full-price sales and on a sign-up given a token value. Everyone agreed the sign-up was not the goal, and nobody wanted to remove it, because the conversion count would collapse and the campaign would look broken.

Our argument is that the worry borrows its logic from count-based bidding. On a value bidding strategy, the conversion count is not what the bidder is chasing, so a large fall in the count can be a small change in what it bids for.

"Secondary goal" is used loosely, and in Google Ads it has a precise meaning. Google says secondary actions are for observation only: they are used for reporting in the 'All conversions' column, but not for bidding, even if the goal they are included in is used for bidding. The one exception is if the secondary action is part of a custom goal, in which case it's used for bidding.

Primary actions, in Google's words, are reported in the 'Conversions' column and used for bidding as long as the standard goal they are part of is used for bidding. So the first check is which of the two the sign-up is. On our reading, a sign-up that is secondary in Google's sense, and not in a custom goal, is not being bid on now, and moving it out of the way changes nothing the bidder uses. The rest of this post is about the common case: a sign-up that sits in the Conversions column beside the sale.

Google says Target ROAS predicts the value of a potential conversion every time a user searches for products or services that you're advertising, and that if the bid strategy determines a search is likely to generate a conversion with high value, it will bid high on that search, and if not, it'll bid low. Google's own table of when to use each strategy separates them on exactly this point: for Maximise conversions and Target CPA, all conversions are treated equally; for Maximise conversion value and Target ROAS, conversions have different values for your business.

What it adds up is the Conversion value column, which Google says shows the sum of conversion values for conversions in your 'Conversions' column. Our reading: a sign-up given a token value contributes its token value to that sum, and no more, however many of them there are.

An illustration, not an account. Take a month with 40 sign-ups valued at £1 each and 10 sales valued at £80 each. The Conversions column reads 50. The Conversion value column reads £840, of which the sign-ups are £40, under 5 per cent.

Remove the sign-ups and, on the same spend, the count falls from 50 to 10, so Cost/conv. is five times higher. Value per conversion rises from £16.80 to £80. Conv. value/cost, the figure Target ROAS steers, falls by under 5 per cent. Two of the three headline columns move by a factor of about five, and the one the strategy is built around barely moves. That is why we judge these campaigns on value per cost against the average target, not on cost per conversion.

The cost is time. Google says Smart Bidding, including target CPA and Target ROAS bidding strategies, considers only conversions reported in the 'Conversions' column when optimising towards your target, and that Google's AI bidding models take time to adapt to any changes that you make to your conversion configuration, at the account level and at the campaign level. If you change your conversion goals at the campaign level, it says you should update your targets gradually over time as the bidding models adjust to newly reported conversions data. Otherwise, you may have unwanted fluctuations in spend.

Google's page on primary and secondary actions gives the standard learning phase that such a change resets. It says Smart Bidding strategies require a standard 7 to 14-day learning phase to gather sufficient data and stabilise ad delivery, and to avoid frequent manual changes to budgets, targets or conversion goals during this period, as these adjustments will reset the learning window.

One thing the token value does carry is count. Google sets a requirement for Target ROAS on Search and Shopping campaigns of at least 15 conversions in the past 30 days at the conversion tracking level. Our practice is to count the sales on their own over the last 30 days before removing anything, because once the sign-ups are gone, the sales are what is left.

This is our reasoning, not Google's. If the price of removing the goal is an adjustment period rather than a cut in value, the decision is about when, not whether. Make the change ahead of the busy season, so the adjustment happens on quieter weeks and the bidder is chasing sales alone by the time demand arrives. Making it in the middle of the peak puts the unsettled weeks on the weeks that matter most.

Our practice, drawn from too few accounts to carry a number. We check whether the sign-up is primary or secondary. We work out its share of the Conversion value column, not of the count. We count sales alone against the 30-day floor. If all three say go, we remove it before the season, leave the target alone for the first weeks and then move it gradually, and judge the result on Conv. value/cost against the average target, never on Cost/conv.