On this page
Agency, rule, or both: the Surplus Test
The agency blames Google’s new rules. Finance blames the agency. Both can be true. Doctor Ads Profit Forensics treats it as a question of dates: which campaigns the change reached, and what else moved. It sits with our other bidding and budget studies.
Call it the Surplus Test. For Target CPA, the surplus is a CPA below the target. For Target ROAS, it is a ROAS above the target. The rule describes that gap moving closer to the target, so each campaign gets three questions.
1. Was it limited by budget before 17 August? The Status column describes this morning, not those days.
2. Was it beating its target before 17 August, and by how much?
3. Did anyone change the budget, target, bid strategy or conversion goals in the window? If yes, two causes can share the result.
Yes, yes and no: a move closer to the target is consistent with the rule, not proof. As our Target CPA study put it, the target is an instruction to Google’s bidding, not a definition of success. Our break-even worksheet sets the right price for Target ROAS. A no on the first or the second means the rule is the wrong explanation.
The dates, from a June rename to a late-October re-check
| Date | What happened, and what it means |
|---|---|
| June 2026 | Labels change for Target CPA and for Target ROAS. The behaviour stays the same, and old names could remain during the transition |
| 5 August 2026 | Google Ads community Q&A explains the update; not proven to be the first announcement |
| 17 August 2026 | Google: rollout starts |
| 27 August 2026 | Google: global rollout completed |
| 26 October 2026 | Our re-check: no newer update, no last-updated date on either page |
For campaigns that are limited by budget and use a target-based bid strategy, Google’s page on changes to target-based bid strategies says: “Starting on 17 August 2026, Google updated its bidding systems”. The FAQ on the change adds: “This global rollout was completed on August 27, 2026, and the new bidding behavior is now fully live across all affected accounts.” Leaving 17 to 27 August out of the comparison is our cut, not a Google window.
What changed in the bidding
About Target CPA bidding states the intent now: “Google Ads will try to keep your cost per conversion equal to the target CPA you set.” Our portfolio study called a target a promise you make to the algorithm.
Who is in, and how a shared budget differs
Google Ads and Search Ads 360: Search, Shopping, Performance Max, Demand Gen and Travel, on Target CPA, Target ROAS, or Target CPC for Demand Gen. Display & Video 360: Demand Gen (FAQ). Out: Manual CPC, Target impression share, App campaigns, video reach or view campaigns. Hotel and Display already steered towards the target. A limited shared budget covers the whole group. Separate budgets under one strategy: only the limited campaigns. A campaign total budget is outside this change.
What did not change
Google says it won’t automatically adjust your targets or budgets, the auction mechanism is unchanged, and campaigns that stay free of a budget limit keep their behaviour. About Smart Bidding learning says “Updates to budgets, ROI targets, or targeting settings take effect in the next auction, but never trigger ‘Learning’ status.” If Learning is cited, ask which change triggered it. One cause on that page is a significant change to the goals, or to the actions the bidding optimises to.
Before 17 August, a surplus was not the usual state
Our published windows end before the change. They show how often a campaign beat its target, not how often it was also limited by budget.
In the Target CPA study, September 2024 to February 2025, 36% of Target CPA campaigns had a CPA more than 20% above target: no surplus for this rule to move. Only 27% of Target ROAS campaigns hit or beat their target, and the median ROAS was 86% of the target.
In smec’s retail data, as PPC Land reported on 22 October 2026, 55% of budget-limited campaigns had been beating their ROAS target, against about a third of campaigns overall. The sample size is not disclosed. Our Performance Max comparison changed direction: mostly September 2024 to February 2025, 71.3% of Performance Max campaigns ran more than 10% above target CPA, against 26.3% of Search; in January to May 2026 the shares were 27.7% and 45.5%.
Lost impression share is the share of eligible impressions a campaign missed, to budget or to rank. In the same smec report, only about 5% of budget-limited campaigns lost that share mainly to budget before the change. After 17 August, budget overtook rank. Our impression-share study found budget took more lost impression share than rank in 14.7% of Search campaign-months (January 2024 to July 2026). That is not a count of limited campaigns.
What Google has not said
Google’s pages describe a mechanism, not the size of any one account’s move. They do not say that every rise since August comes from this change, or that a campaign missing its target will now hit it. “The update ended bid suppression”, campaigns bidding below the target on purpose, is other people’s reading, not Google’s wording.
What other datasets found
smec, in the same PPC Land report: the median budget-limited campaign had beaten its ROAS target by about 20% in data back to 1 June, and that surplus was about zero by 31 August. Equeco, on 19 October 2026, could not identify which of its Search campaigns qualified for the change, so it could not compare affected and unaffected campaigns directly; it measured campaigns with consistent volume before the update. Against the week of 10 August, its median campaign cost per click peaked at +5.1% in the week of 31 August and was +0.9% in the week of 28 September. Cost per click is not cost per conversion.
A rise that starts in October needs an October cause, or a budget limit that began later. In Equeco’s Search campaigns the median CPC shift had faded by late September. PPC Hero, Search Engine Journal, Optmyzr and Google’s help pages state the rule. None separates the agency from the rule with dates.
What to check in your account
Google dates the rollout 17 to 27 August 2026; we compare a window that ends before 17 August with one that starts after 27 August. Use the same length for both. Do not judge the later window on its first short cycle. Wait one or two full conversion cycles (target adjustments).
- List the campaigns in scope. Add Bid strategy type. Keep Target CPA, Target ROAS and Demand Gen on Target CPC. Ignore today’s Status column.
- Check the budget limit for 1 to 16 August 2026. Take each day’s budget from change history. Limited by budget means your average daily budget is lower than the recommended amount to capture all available impressions and clicks. Cost near it is a sign, not the status. A day can reach twice the average daily budget (overdelivery). Shared budget: the whole group. Separate budgets: only the limited campaigns. Campaign total budget: outside this change. Cost well under the budget is not the status (underspend study).
- Measure the surplus on both windows: Avg. target CPA beside Cost / conv., Avg. target ROAS beside Conv. value / cost, and average CPC beside the CPC target on Demand Gen. The average is traffic-weighted and includes target edits, device adjustments and ad group targets (About Target CPA). Use at least 30 conversions.
- Read the later window. A CPA rising from below the target, or a ROAS falling from above it, is consistent with the rule. A CPA above the target, or a ROAS below it, needs another cause.
- Compare channels on Performance Max and Demand Gen. In both windows, compare each channel’s share of cost and of conversions (the change, the FAQ).
- Read change history in the Campaigns menu, for 1 August to 26 October 2026: budget, strategy, target and conversions. Change history misses some account-level edits and some edits by Google representatives in a consultation. An empty history is not proof. A limit that starts after the rollout is a separate case.
- Check conversions. Goals, then Conversions, then Summary shows today’s settings. Match primary actions and the counting method, once or every time, across the two windows in change history.
The question to send the agency
“For each campaign whose CPA or ROAS moved after 17 August 2026, send: (1) daily cost against the budget from change history, for example 1 to 16 August 2026. Cost near the budget is not Limited by budget, a day can reach twice the average daily budget, and a shared budget is judged for the group. (2) CPA, ROAS or average CPC against its target, on two equal windows, one ending before 17 August and one starting after 27 August. (3) change history from 1 August to 26 October 2026. Some account-level edits, and some edits by Google representatives in a consultation, are missing, so a blank history is not proof. Say where this is consistent with the move.”
If the answer blames Google’s bidding for every campaign, ask the agency to show the three checks. See our questions to ask your agency.
The line for the board
“Google’s rollout began on 17 August 2026 and finished on 27 August. Where a campaign was budget-limited before 17 August, was beating its target (CPA below it, or ROAS above it), and budget, target, strategy and conversion goals were unchanged, a move closer to the target is consistent with the rule. Campaigns [A, B] met all three. Campaigns [C, D] did not. Source: Google Ads Help, checked 26 October 2026.”
When this does not apply
It was not limited by budget before the change, and it stayed that way. The rise needs a dated cause. Raising the budget because of August spends money the rule did not touch.
It was already missing its target (CPA above it, or ROAS below it). Google does not promise it will now land on the target.
The later result is worse than the target. Check for another cause as well: Google describes a surplus shrinking and has not said whether the rule can push a result past the target.
Something else moved. An edit to budget, target, strategy or conversion goals means two causes can share the result.
When to call an independent check
Call one when the agency’s story and the screens disagree, or when budgets, target history and conversion settings cannot be rebuilt for the dates. Splitting what Google changed from what management changed is the first page of a Profit Forensics verdict. The honest result can be that the rule fits and the target is the wrong price, that the rule does not fit, or a clean bill with nothing in the management to fix.
Researcher’s take
I would not accuse the agency, or close because a post said the rules changed. The move is consistent with the rule only where the campaign was limited by budget, beating its target (a CPA below it, or a ROAS above it), and untouched on budget, target, strategy and conversion goals. If one condition fails, ask the agency to show the campaign data.
Igor Ivitskiy, Doctor Ads Profit Forensics
FAQ
Did the agency get worse, or did Google change the rules?
Not from the cost alone. The rule fits a campaign only if it was limited by budget before 17 August, was beating its target and kept the same budget, target, strategy and conversion goals. If one of those changed, two causes can share the result.
Our CPA ran well under target and rose to it after 17 August. Is that the change?
Consistent with the rule if it was budget-limited, beating its target, and unchanged on budget, target, strategy and conversion goals. Check budget history, the surplus and later edits first.
Will the 17 August change pull our campaigns back to target?
Only where all three conditions hold. Google does not guarantee a landing on the target. The rollout finished on 27 August 2026 (FAQ). Target adjustments: “it’s recommended to wait 1-2 cycles before assessing performance” (1-2 cycles).
How we know
Google Ads Help, checked 26 October 2026: Changes to target-based bid strategies; FAQ: changes to Target-based bid strategies; About Target CPA bidding; About bid strategy statuses; How to make target adjustments; About Smart Bidding learning; Limited by budget; Overdelivery; Change history. The Google Ads community Q&A of 5 August 2026 only dates the announcement.
Our published figures above all predate the change: the Target CPA study, the impression-share study and the Performance Max comparison. They describe a managed portfolio, not a random sample. The smec and Equeco figures are their data, not ours.
Opinion, not measurement: the Surplus Test, the order of the checks, the question and the board line are Igor Ivitskiy’s recommendations.
No new portfolio data in this article: it combines Google’s documentation (checked 26 October 2026) with what our audits look for. Where a figure comes from our published studies, the study is linked.
The split between the platform’s change and the account’s own edits is where a Profit Forensics verdict starts.