Google Ads since August 17: your CPA or ROAS target is now an instruction, not a ceiling

Since August 17, 2026, a Google Ads campaign limited by its budget and driven by a target cost per acquisition or a target return on ad spend optimizes toward the target you entered, and no further. That safety ceiling has become an instruction. A business whose target does not come from its gross margin pays for the change.

On this page

Who announced it, and when

Google published the change on June 15, 2026 in the Google Ads help centre, in a batch of three announcements covering bidding and budgets. The target adjustment tool appeared in accounts on July 6, along with a notice sent to every advertiser that had run a budget-limited campaign in the previous twelve months. The new behaviour took effect on August 17, rolling out in waves over several weeks rather than in a single day.

The spokesperson was Ginny Marvin, Google Ads product liaison, who answered advertisers on LinkedIn all summer and published a question-and-answer session on August 6. Her most useful formulation fits in one sentence: in a budget-constrained campaign, the target becomes your efficiency lever. The budget controls spend, the target controls return, and the two no longer contaminate each other.

The industry reaction was sharp, because a share of account managers were deliberately using a loose target to let the system explore and find cheaper conversions. To them, Google is removing a lever that worked. To Google, it is fixing an inconsistency: the same target was supposed to produce the same efficiency whether the daily budget was $50 or $500, and it did not.

Who is affected, and who is not

The scope is precise, and most small and mid-sized business accounts fall inside it entirely, because their budgets are almost always capped. That is an observation from our engagements: on the accounts we take over, the main campaign is limited by budget more often than not.

Two conditions have to be met: the campaign carries a target, and its budget limits it. A campaign with a target whose budget never runs out does not change. A campaign on "maximize conversions" with no target does not change either. Portfolio strategies and shared budgets are affected at the portfolio or shared budget level, and that is where the adjustment has to be made, never campaign by campaign.

What it does to an account, in numbers

Take an online store. Gross margin of 45%. Google Ads budget capped at $3,000 a month. Target return on ad spend set at 300% two years ago and never revisited. For the last six months the campaign has actually been delivering 520%, because the system, constrained by the budget, was concentrating spend on the cheapest auctions.

At 520%, that $3,000 produced $15,600 in sales, roughly $7,000 of gross margin and $4,000 of contribution once the advertising was paid for. If the campaign converges toward its 300% target, the same $3,000 produces $9,000 in sales, $4,050 of margin and $1,050 of contribution. Nothing changed in the account. Nobody touched anything. Monthly contribution was divided by close to four.

This is the point Google's communication walks around. The promise is being able to grow with confidence at the target you entered. That promise is true, but it only materializes if you raise the budget. On a fixed budget, the return converges downward and the spend stays the same: the update cannot give you anything, it can only cost you. The gain exists for the advertiser who has demand above their cap and the means to go get it.

What Google did not say: how the campaign climbs back toward the target

The documentation explains the outcome, not the mechanism. An advertiser put the question to Google directly in August: for a campaign delivering $5 to climb back to $10, will the system bid higher on the same conversions, or spend on search terms it used to leave aside? The question went unanswered in public. The two hypotheses do not cost the same. The first makes you pay more for the same volume. The second buys you lower-quality volume, and it is the one that hurts B2B accounts, where a poorly qualified enquiry costs sales time after the click.

Until Google documents the mechanism, two report lines settle the question for your account: average cost per click and the search terms report, compared with the six weeks before August 17. If cost per click rises and the terms do not move, it is the first hypothesis. If the terms widen, it is the second, and the answer is not in the target, it is in the exclusions.

Our position: the update takes nothing from those who had done the math

The disagreement of this summer was never about Google's documentation. It is about what a target represents. To Google, it is the return you are asking for. To a large share of managers, it had become one control among others inside a capped campaign: you left $50 in the field because the campaign was delivering $35, because leadership was satisfied, and because nothing forced you to touch it. Those are two uses of the same setting, and on August 17 they were forcibly glued back together.

Two readings have been competing since June. The first says Google is taking back free efficiency and calling it a feature. The second says Google is clarifying a control that was ambiguous. Both are accurate. What the debate is missing is saying who each one applies to.

The reported return has never been the real return. We have been writing it for a long time: a ROAS of 4 is a loss for a business with a 20% gross margin, and the break-even threshold of a Google Ads campaign is set before optimization, from the margin. An account managed that way already carries a target that follows from a calculation. For that account, August 17 does not exist: the target was already the instruction, and the budget was already the volume lever.

The account that suffers is the one whose target was a comfort number. Someone put 300% in because the previous vendor had, or $50 because that was the maximum acceptable one day in 2024, and every month a return above target was celebrated without anyone asking what the target meant. That account is not losing efficiency. It is losing an illusion, and Google has just sent it the invoice for never having done the math.

We recognize a real loss for one precise category of advertisers: those who deliberately left a loose target so the system would explore, on a small budget, pockets of cheap conversions nobody could have found by hand. It was a strategy that worked. But it was also a strategy that delegated the very definition of profitability to the algorithm, and what you do not define you can neither defend nor bill for. That delegation has just ended. The prospecting work becomes human work again, grounded in the account's own data, and it is worth what you pay for it.

Four situations, four decisions

The gap between the target and the actual return does not tell you what to do. You first have to know where the number came from, then whether the budget can move. Those two questions give four situations.

The fourth row deserves a note, because it covers most small and mid-sized business accounts. Google's tool calculates no recommendation for a campaign too thin in conversions, and an account closing thirty sales a month does not have the material for a tight target to work. On those accounts, our practice is to let the budget do the ceiling work and to measure profitability outside the system, in the monthly report, against the margin.

How to execute without misreading it

The order matters, and so does the calendar. August 17 falls ten weeks before the fourth quarter. For an online store, the first defensible reading arrives when the holiday budgets are already committed. For a B2B business with a three-week sales cycle, one or two conversion cycles does not mean a few days: it means four to six weeks before the bid strategy report reflects anything other than the system relearning.

1. Pull the list, and widen it

Every campaign that has carried the limited-by-budget state over twelve months, not only those carrying it today. Keep only those on target CPA, target ROAS, or target CPC in Demand Gen. Then draw the distinction: a campaign that capped for two days in April does not deserve the same treatment as one that has spent its full budget every month for a year. The second one is the real subject.

2. Put the actual next to the target, and rank by money at stake

Thirty days for a high-volume campaign, sixty to ninety for a thin one. Then rank by the gap multiplied by spend, not by the size of the gap. A twofold gap on $400 a month weighs less than a 20% gap on $8,000.

3. Ask the margin question before touching the number

Recalculate what the business can pay per acquisition, or the minimum return that covers gross margin and the expected contribution. That calculation decides, not the report from the last three months. A target that survives that test was probably deliberate. A target that does not was inherited.

And sometimes the target is not the problem. A campaign can be limited simply because the budget is too small for the demand available, or because its keyword coverage is too broad for that budget. In those cases, tightening the exclusions or revisiting the scope weighs more than any number in the target field. Changing the target is one option in the review, not its automatic answer.

4. Apply one change at a time, and write it down

Old target, new target, date, one line of reasoning. Four weeks from now, that is what separates a diagnosis from a guess. Several managers choose to move in steps rather than in a single jump, and to look at each step before the next. Google says one clean adjustment is enough. Both positions can be right depending on the account.

5. Judge nothing before the window closes

First three days: look, touch nothing. Through day fourteen: read the bid strategy report, ignore the performance planner, which Google itself announced as unreliable between August 17 and 31. Weeks two to four: correct what was set too quickly, one change at a time. After that, this check becomes a permanent line in the report, because the gap reopens as soon as the market moves, and a reopened gap is no longer free efficiency: it is a target that has aged.

To verify before concluding that the update hurt your results

Frequently asked questions about the August 17, 2026 Google Ads update