Quick answer: Google began a gradual rollout around August 17, 2026 that removes target overperformance in Smart Bidding — budget-limited Target CPA and Target ROAS campaigns will now aim at the stated target rather than beating it. If your reporting looks worse, audit target-versus-actual gaps, conversion quality and business thresholds before changing any bids or budgets.
What actually changed
For years, a budget-limited Target CPA or Target ROAS campaign often delivered results better than its stated target. The reason was structural: the budget ran out before the target became the binding constraint, so the system bought only the cheapest available conversions. Many advertisers came to treat that overperformance as the campaign's true performance.
Google Ads Help now describes bidding as aiming at the target you set, with the previous overperformance behaviour being removed on a gradual rollout that began around August 17, 2026. The practical consequence is that in affected campaigns, actual CPA can rise toward your stated target and actual ROAS can fall toward it — with no change in your settings and no decline in campaign health.
Note what Google did not publish: a completion date, a list of affected accounts, or an expected magnitude of change. Treat any specific percentage claim you read elsewhere as unverified.
Step 1: filter to the campaigns this can affect
Do not audit the whole account. Build a filtered view of campaigns that meet all three conditions:
- Bid strategy is Target CPA or Target ROAS (including portfolio strategies using them).
- Status shows or recently showed "Limited by budget".
- Actual CPA or ROAS was consistently better than the stated target before mid-August.
Campaigns failing any condition are out of scope for this specific change. Auditing them anyway is how you end up "fixing" campaigns that were never affected.
Step 2: measure target-to-actual gaps, not month-over-month
This is the single most important reframing. For each in-scope campaign, record the stated target, the actual result for a pre-rollout window, and the actual result for a post-rollout window of comparable length and conversion volume.
Then ask a different question than usual: not "did performance decline" but "how large was the gap between target and actual before, and how large is it now". A campaign whose CPA was 40% below target and is now 5% below target has not deteriorated — it has stopped subsidising a target you set too generously. The number that changed is the gap, and the gap was never a performance metric.
Keep the windows honest: same length, comparable weekday composition, and enough conversions to be meaningful for your volume. Also record the rollout status per campaign where you can, because a gradual rollout means campaigns may have transitioned on different dates.
Step 3: validate conversion quality before touching anything
Smart Bidding optimizes toward the conversions you report. If those are wrong, everything downstream is wrong. Verify:
- Which actions are primary. Only actions that represent real business value belong in the bidding set; secondary actions should be observation-only.
- Duplicate counting. A single lead recorded by both a form submit and a thank-you pageview halves your apparent CPA and corrupts the target.
- Conversion values. Target ROAS is meaningless without accurate values. Check that they reflect margin or realistic revenue rather than placeholders.
- Attribution and lookback settings. Note the model and window; changing them mid-audit invalidates the comparison.
- Lead quality downstream. Reconcile reported conversions against qualified leads or closed revenue in your CRM. A stable CPA hiding a collapse in lead quality is worse than a rise in CPA.
- Import and consent status. Confirm offline imports are still landing and consent-mode changes have not silently reduced recorded conversions.
Step 4: restate the business threshold
Your target should be derived from economics, not from the number the account happened to be producing. Work it out explicitly: average order value or lead value, close rate, gross margin, and the maximum acquisition cost that still leaves the transaction worthwhile. That figure is your target ceiling.
Compare it to your current setting. Three outcomes follow. If your stated target is well above your true ceiling, the overperformance was hiding a target that was never defensible — tighten it deliberately. If your stated target matches your ceiling, leave it and accept that actual results will now sit near it. And if actual results now exceed your true ceiling, that is a genuine profitability problem requiring a decision about volume, not a reporting artefact.
Step 5: act, in one direction at a time
Four paths cover almost every in-scope campaign:
Hold
Target reflects the business threshold, actual sits at or below it, volume is acceptable. Change nothing and update the reporting narrative so stakeholders are not told a story about decline.
Tighten the target
Actual now exceeds your true ceiling. Reduce the target in modest increments — large jumps stall delivery — and allow a full conversion cycle between adjustments before judging the result.
Raise the budget
Actual is comfortably inside your ceiling and the campaign is still budget-limited with incremental demand available. This is now a clearer trade than before: additional budget buys volume at approximately the target rather than dragging efficiency somewhere unpredictable.
Restructure
Conversion volume is too low for the target to be learnable, or several unlike offers share one campaign. Consolidate for signal density or split by economics — bidding strategy changes cannot fix a structural problem.
Whichever path you take, change one variable per campaign per cycle and log the date. Our marketing reporting setups keep target-versus-actual visible alongside spend precisely so these decisions do not get made from a single blended dashboard number.
Target CPA and Target ROAS specifics
For Target CPA, the risk after this change is over-correction: cutting the target to restore an old CPA figure typically buys a volume collapse. Judge it against your ceiling, not against August.
For Target ROAS, the value data is the whole strategy. If values are stale, uniform, or set to placeholder amounts, the target expresses nothing about profitability and the campaign will optimize toward the wrong outcome regardless of what you set. Fix values first; adjust the target second.
What to tell stakeholders
Report the change explicitly: a confirmed platform rollout altered how budget-limited Smart Bidding campaigns treat targets, and the effect on our reported CPA or ROAS is X, measured over these windows. State clearly that this is a platform behaviour change, not a decline in campaign management — and state equally clearly if the audit revealed a target that was never economically defensible. Both are useful; hiding either is not.
If you want a second pair of eyes on the target-versus-actual work, our PPC advertising team and our Vancouver PPC agency practice run these audits regularly, and you can contact us for a free consultation.
The accounts that come out of this rollout in better shape are the ones that used it to check whether their targets ever matched their economics.
