From 17 August, Google is changing how Smart Bidding handles Target CPA and Target ROAS campaigns. The specific behaviour being removed is intentional overperformance - where the algorithm was allowed to push spend further in pursuit of results that exceeded the stated target. After the deadline, Smart Bidding will operate strictly in line with the target you set. No more quietly outperforming to justify additional spend. If your targets do not reflect where you actually want the algorithm to operate, you have a short window to fix them.
What Overperformance Actually Meant in Practice
Target CPA and Target ROAS have always been instructions to the algorithm, not hard constraints. Smart Bidding would use these targets as a central reference point but could bid above or below depending on the auction signals it was reading. Overperformance meant that if your tCPA was set at £60, the system might actively pursue conversions at £45 if it judged the opportunity worth taking - spending more volume in the process.
For many accounts, this created a useful buffer. Advertisers would set a slightly conservative target, allow the algorithm room to operate, and benefit from a blended CPA that came in below the stated threshold. It was a common enough tactic that some accounts were structured around it. The problem is that after 17 August, that buffer disappears. The algorithm will anchor to your target much more precisely, and if that target is set too conservatively, you could see volume drop as the system declines auctions it would previously have entered.
Which Campaigns to Audit First
Not every account will feel this equally. The campaigns most exposed are those where the actual achieved CPA or ROAS has consistently come in significantly better than the stated target. Pull your campaign-level performance data for the last 60 to 90 days. If your tCPA is £80 but your actual CPA is running at £52, you have been benefitting from overperformance. That gap is where the change will bite.
The same logic applies to tROAS. If your target is 300% but the campaign has been delivering 420%, the algorithm has been operating in overperformance mode. Once that stops, you either need to raise the tROAS target closer to your actual ambition, or accept that the campaign will likely scale back. Performance Max campaigns using Smart Bidding targets are included in this change, so do not limit your audit to standard search and shopping.
Flag campaigns in the following categories as priority reviews: lead gen campaigns with a significant gap between target and actual CPA, ecommerce campaigns where actual ROAS consistently exceeds the stated target, and any campaign where budget was previously being fully spent despite what appeared to be a conservative target. If the budget was clearing comfortably, overperformance was probably part of why.
How to Update Your Targets Correctly
The adjustment required is straightforward in principle, but needs care in execution. You need to move your stated targets closer to where the algorithm has actually been operating. For a tCPA campaign running at £52 against a £80 target, a revised target somewhere in the £55 to £60 range is a reasonable starting point - close enough to signal real intent to the algorithm without setting something you cannot sustain if conditions change.
Do not make aggressive target changes across all campaigns simultaneously. Smart Bidding goes through a learning period when targets shift materially, and changing multiple campaigns at once can create instability across the account. Prioritise the campaigns with the widest gap between target and actual performance, make the update, monitor for five to seven days, then move to the next group. Document what each target was before you changed it - you will want that reference if performance moves in an unexpected direction after 17 August.
For campaigns where you genuinely cannot decide what the right target should be, use a portfolio bidding strategy. Portfolio targets give you a single point of control across multiple campaigns and can provide more flexibility as you work out where the algorithm settles post-change. It is also easier to adjust one portfolio target than to chase individual campaign settings under time pressure.
The Broader Implication for How Targets Are Set
This change is a signal about where Google wants Smart Bidding to go. The previous overperformance behaviour was, in a sense, an unofficial mechanism that allowed the algorithm some discretion beyond the stated goal. Removing it makes the system more literal. Your target is now closer to a genuine instruction rather than a starting point for negotiation.
That is actually a more honest relationship between advertiser intent and algorithm behaviour - but it requires advertisers to set targets with more precision than many have been used to. The habit of setting a conservative target and hoping the algorithm does better was always a workaround, not a strategy. From 17 August, the case for setting targets that genuinely reflect your cost per acquisition or return goals becomes stronger. If your business can sustain a £60 CPA, set £60. If you need 350% ROAS to make the margin work, set 350%.
It also reinforces why conversion tracking quality matters so much. Smart Bidding is only as good as the signals it is optimising towards. If your conversion tracking is incomplete - missing offline conversions, undercounting due to consent mode gaps, or attributing value incorrectly - then tightening the relationship between target and algorithm behaviour amplifies those underlying issues. A precise target against flawed conversion data is not an improvement. Audit your conversion setup alongside your bidding targets before the deadline.
What to Watch After 17 August
Even if you update your targets correctly, expect some turbulence in the week or two after the change goes live. The algorithm will be recalibrating across many accounts simultaneously, not just yours. Impression share, click volume and conversion rates may all shift during this period. Avoid making reactive changes based on two or three days of data.
Set up custom alerts in Google Ads for significant CPA or conversion volume changes so you are not checking dashboards constantly. Pull a week-on-week comparison from your analytics setup rather than day-on-day, which will help filter out daily noise. If you are running experiments or campaign tests, pause them during this period so you are not trying to read two variables at once.
The accounts that will handle this best are those with clean conversion tracking, targets that already reflect actual business economics, and a team that has documented the state of their campaigns before the change. That documentation is not optional - if something does move unexpectedly, you need a clear baseline to compare against rather than trying to reconstruct what things looked like from memory.