Your Target CPA Is Now a Target

For years, a budget-limited Google Ads campaign on Target CPA did something nobody wrote down. You typed £35. The system found the cheapest auctions it could reach inside your budget, and delivered £22. You reported £22. Nobody asked why the number you asked for and the number you got were different, because the difference was in your favour.

From 17 August 2026, that stopped. Budget-limited campaigns on Target CPA, Target ROAS, and Target CPC for Demand Gen now optimise more literally toward the target you typed. The £22 drifts back toward £35.

Nothing broke. The system is doing what it was told. What has been exposed is that a great many targets were never targets at all — they were ceilings nobody expected to touch, and the gap between what was typed and what was delivered was carrying weight that nobody had accounted for.

What actually changed

Google's wording is that campaigns limited by budget "will more consistently perform toward your bid target, including when you make budget adjustments." The scope is wide: Search, Shopping, Performance Max, Demand Gen and Travel, through the interface, Search Ads 360, DV360, Editor and the API.

The mechanism is worth being precise about, because the reporting has mostly described the symptom. Under a binding budget, Smart Bidding previously had a choice: it could enter every auction it thought would convert at or under the target, or it could enter only the cheapest of them. With the budget as the binding constraint, it took the cheap ones. That produced a flattering CPA and a conversion count limited to the cheap end of the market.

Now the budget is spent toward the target rather than under it. More expensive auctions come into range. The average price paid rises toward the number you typed.

The arithmetic nobody enjoys

A fixed budget divided by a price is a quantity. That is the whole of the risk here, and it needs no modelling.

Ten thousand pounds at £22 buys 455 conversions. The same ten thousand at £35 buys 286. That is 37% fewer, for the same money, with no change in your product, your landing pages or your market.

There is one way that arithmetic is not the end of the story, and it is the question the whole change turns on: are the dearer auctions incremental?

If the more expensive impressions reach buyers the cheap ones never did — a different segment, a later stage, a competitor's customer — then 286 conversions at £35 may be worth more to the business than 455 at £22, because more of them are conversions that would not have happened anyway. If the dearer auctions are simply the same buyers bought at a worse price, then you have just paid 59% more per unit for nothing.

A bidding setting cannot tell you which. Only a holdout can.

What to do this week

Re-derive the target from margin, not from last month's actuals. If your reported CPA was £22 because the system was cherry-picking, then £22 was never a business number — it was an artefact of a constraint. Work out what you can afford to pay for a conversion at your contribution margin and your payback period, and type that.

Then decide whether you are buying volume or price. These are now separable in a way they were not before. If the target is right and the budget is binding, you are choosing to buy fewer, dearer conversions. That is a legitimate choice — but it should be a choice, not a surprise in a monthly report.

Run a geo holdout before you conclude anything. The one question that matters — whether the extra spend reaches anybody new — is an incrementality question, and it is answerable in three or four weeks with a split by region. Everything else is inference from a number that changed for a known reason.

And check what your reporting is anchored to. Any dashboard, bonus or agency SLA written against "CPA under £25" is now measuring a different world than the one it was written in.

Why this is a measurement story and not a bidding one

The reason this change stings is not that a setting moved. It is that a number a lot of teams were judged on turns out to have been describing something other than what everyone believed.

That is an old effect with a name. When a measure becomes a target, it stops being a good measure — and the corollary, less often stated, is that a measure which was never really a target can be quietly holding a lot of weight. Reported CPA was doing exactly that: it looked like a measure of efficiency and was substantially a measure of how binding your budget was.

The fix is the same as it always is. Steer by more than one number, keep one of them out of the bonus, and read direction over months rather than a level on a date.

Conclusions

A target that was never enforced was never a target. For a lot of accounts, the number in the Target CPA box was aspirational and the number in the report was an artefact of a binding budget. Only one of those was ever a business decision, and from 17 August it is the one that governs.

At a fixed budget, price and volume now trade off visibly. £10,000 at £22 is 455 conversions; at £35 it is 286. That is division, not a forecast, and it will show up in the monthly report whether or not anybody plans for it.

The only question worth answering is incrementality. Whether 286 dearer conversions beat 455 cheap ones depends entirely on whether the dearer auctions reach anybody new. A geo holdout answers it in a month. No amount of tuning inside the platform will.

And check what is anchored to the old number. Dashboards, bonuses and agency agreements written against a CPA that the system used to beat by accident are now measuring a different world. Re-anchor them deliberately, before somebody is judged against a figure that no longer means what it did in July.