AI + Paid Media
Google Wants a Looser Target. Profit Still Has to Approve the Trade.

Updated: October 28, 2026
A recommendation to raise target CPA or lower target ROAS is a proposal to buy more auction access at a weaker efficiency constraint—not proof that the extra volume will be profitable. Translate the proposed multiplier into dollars, confirm conversion quality and capacity, test one bounded change, and define the rollback rule before applying it.
What Google added
Google Ads API v25.2 added two Search recommendation types for campaigns whose performance is limited by bidding targets: one can recommend raising target CPA and another can recommend lowering target ROAS. The response can include a recommended multiplier and the current average target. The release shipped September 23, 2026, and the fields remain part of the current API line.
The system is identifying auction eligibility headroom. It is not certifying contribution margin, lead quality, cash flow or operational capacity.
Translate the multiplier into a business decision
| Platform proposal | Business translation | Approval question |
|---|---|---|
| Raise target CPA | Pay more per conversion to enter more auctions | What is the maximum profitable acquisition cost by customer type? |
| Lower target ROAS | Accept less reported value per ad dollar | Does contribution profit still clear the hurdle? |
| More eligible traffic | Potentially more volume and more weak demand | Can sales, inventory and service absorb it? |
| Recommended multiplier | A proposed change, not a guarantee | What evidence would cause a rollback? |
Diagnose the actual constraint first
Check budget status, lost impression share, conversion volume, learning state, audience or query quality, landing-page capacity and offline signal health. A low bid target can constrain auctions, but a broken conversion definition can also make the target look artificially difficult. A budget increase and a target relaxation solve different problems.
Compare this decision with MTC’s raise-budget recommendation review. One changes the spending ceiling; the other changes the efficiency constraint. Do not approve both simultaneously unless the test is explicitly designed to measure their combined effect.
Build the marginal economics
- Start with the current target and actual CPA or ROAS.
- Apply Google’s proposed multiplier to estimate the new constraint.
- Convert platform value to gross margin or contribution profit.
- Apply lead-to-sale, refund, cancellation and repeat-purchase adjustments.
- Estimate the additional volume required to offset weaker unit economics.
- Confirm cash, inventory and sales capacity.
Run the model separately for new customers, existing customers and materially different products. A blended account target can hide a segment that becomes unprofitable first.
Test one controlled step
Use an experiment when available, or change a limited campaign group while holding conversion actions, attribution settings, creative and landing pages stable. A smaller movement than the full recommendation may be appropriate. Record the change time and allow for conversion lag before declaring a winner.
Primary evaluation should include qualified conversions, realized value and contribution profit. Auction access, clicks and platform conversions explain delivery; they do not replace the commercial outcome.
Set rollback rules before launch
- Maximum acceptable qualified CPA or minimum contribution ROAS.
- Minimum conversion-quality or sales-acceptance rate.
- Spend or time boundary for the first review.
- Guardrail for refunds, cancellations or low-margin mix.
- Rule for returning to the prior target.
- Owner who can approve a second step.
Use a decision memo, not a screenshot
Capture the recommendation date, campaign, current target, proposed multiplier, observed headroom, business-unit economics and final decision. Add the reason for rejecting the recommendation when the target is intentionally strict. This prevents the same suggestion from being reconsidered without new evidence and gives future reviewers a record of what the platform knew at the time.
After the test, compare actual marginal results with the pre-launch model. If the campaign gained volume at a lower qualified close rate, update the model and offline conversion feedback. If it produced profitable scale, document the range in which it worked instead of treating the new target as permanent.
Common failure modes
Optimizing a blended conversion
A looser target can buy more low-value actions when every conversion shares one value. Separate primary outcomes and pass reliable values.
Judging before lag resolves
Target changes can alter delivery before late conversions arrive. Use the real sales-cycle delay.
Making the recommendation the hypothesis
The useful hypothesis is commercial: relaxing the constraint will produce additional profitable customers. The recommendation only suggests how to test it.
Frequently asked questions
Does a “performance bid too low” recommendation mean the campaign is broken?
No. It means the current target may restrict auction participation under Google’s model. The target may still be correct for the business.
Should target CPA and budget change together?
Usually not in a diagnostic test. Changing both makes it difficult to identify which control caused the result.
What is the best success metric?
Incremental qualified contribution profit is stronger than platform volume alone.
