Export the baseline
Save the target, budget status, actual CPA or ROAS, conversion delay, and conversion volume before changing the setting.
Stan Consulting note | Google Ads target control
The target you typed is about to become the result you asked for.
A stale setting is not harmless account history when the platform is told to follow it more closely.
Target CPA$10actual $5↓tighten down
Target ROAS400%actual 550%↑tighten up
The direct answer
Google says that starting August 17, affected campaigns limited by budget will optimize more consistently toward the target written in the settings. Google will not adjust that target for the advertiser.
That turns an old number into an active decision.
A Target CPA campaign can show a $10 target while quietly delivering $5 acquisitions. A Target ROAS campaign can show a 400 percent target while quietly delivering 550 percent. The account looks healthy because the actual result is better than the setting.
Then the platform follows the setting more closely.
The gap you enjoyed was not the target. It was the part the target allowed the system to give away.
The contradiction
The same instruction is correct for ROAS and wrong for CPA.
| Strategy | Setting | Actual | To defend the better actual |
|---|---|---|---|
| Target CPA | $10 | $5 | Lower the target toward $5. |
| Target ROAS | 400% | 550% | Raise the target toward 550%. |
“Raise the target” sounds decisive. It is also incomplete enough to damage a CPA campaign.
What the setting really is
A target is permission.
It tells the system how much cost or how little return the account is allowed to accept while it uses the available budget.
When the target is loose, the account may outperform it. That outperformance does not prove the target was wise. It proves the market, conversion signal, auction mix, or historical model produced a better result despite the instruction.
The August change removes some of that comfort. A forgotten target can pull the result toward a number nobody would defend in a budget meeting.
The account check
Do not edit from memory.
Check the signal
A precise target still optimizes the wrong outcome when the account rewards weak leads, duplicate conversions, returns, or revenue without margin.
Move the right direction
Lower Target CPA toward a better actual CPA. Raise Target ROAS toward a better actual ROAS.
Wait long enough
Give the change one to two conversion cycles. The first day is movement, not evidence.
The operating judgment
The platform did not choose the bad number.
It found it in the account.
That is the uncomfortable part. Automation makes the old decision more consistent. It does not make the old decision correct.
A target that nobody owns becomes a silent transfer of authority from the business to the platform. The platform will do exactly what it was allowed to do, then report the result as optimization.
The consequence
A stale target is a live permission.
On August 17, the account does not become less automated.
It becomes more obedient.