Your budget-limited campaigns have been quietly outperforming their targets. On August 17, that stops. If you haven't audited the gap between your stated Target CPA and your actual CPA, you're about to pay more for the same volume, and your next pipeline review will surface the variance before you do.
Google's documentation frames this as a predictability improvement: campaigns will now optimize toward the number you entered, not some better number the algorithm found on its own. That's true. It's also a margin event for any team that built forecasts on performance the system was never contractually obligated to deliver.
The Mechanics, Without the Jargon
Here's what's actually happening. A campaign marked "Limited by budget" using Target CPA or Target ROAS has historically had room to hunt for cheaper conversions than you requested. You set a $50 Target CPA, the system found pockets at $32, and you pocketed the difference as unearned efficiency. Smart Bidding was doing discovery work you didn't ask for, and you weren't paying for it.
Starting , Google is tightening that leeway. Budget-constrained campaigns will optimize closer to your stated target. A $50 Target CPA will deliver something much closer to $50, not $32. The algorithm stops overdelivering.
This applies to Search, Shopping, Performance Max, Demand Gen, and Travel campaigns that are budget-limited. Campaigns with headroom in their daily budgets won't change. App, Video Reach, and Video View campaigns are excluded. Hotel and Display already work this way.
The critical point: Google does not adjust your targets for you. The responsibility to act is entirely yours.
Why This Is a Forecast Problem, Not Just a Media Problem
Most marketing teams treat paid media efficiency as a channel metric. It lives in a dashboard, gets reviewed weekly, and rolls up into a blended CAC number that finance sees quarterly. That's fine until the efficiency you've been reporting was never locked in.
If your actual CPA has been running 30% below your stated target, your CAC payback model is built on a number you don't control. When that gap closes on August 17, your cost per opportunity rises, your pipeline coverage math shifts, and your CFO asks why Q3 acquisition costs jumped without a corresponding increase in volume.
This is the kind of variance that erodes trust between marketing and finance. Not because anyone lied, but because the system was doing something nobody explicitly asked it to do, and nobody flagged the dependency.
The Audit That Matters
Google released a Bid Target Adjustment Tool on July 6. Use it. But the tool only tells you which campaigns are affected and what your actual performance has been. It doesn't tell you what to do about it.
Here's the decision framework:
Step One: Identify the Gap
Pull every budget-limited campaign using Target CPA or Target ROAS. Compare your stated target to your trailing 30-day actual. If your Target CPA is $50 and your actual CPA is $35, you have a 30% gap that's about to close.
Step Two: Decide What You're Optimizing For
If you want to maintain your current efficiency, lower your target to match your actual performance. That $50 target becomes $35. The system will now optimize toward the number you were actually getting.
If you want to maintain your current volume and you have budget flexibility, keep your target where it is and increase your daily budget so the campaign is no longer constrained. The change only affects budget-limited campaigns.
If you want to scale volume and you're willing to pay more per conversion, keep your target where it is and accept the higher CPA. Just make sure your CAC payback model still works at the new number.
Step Three: Reforecast
This is the step most teams skip. If your blended CAC assumption was built on campaigns that were overperforming, your Q3 and Q4 forecasts need to be updated before August 17, not after. Run the sensitivity: what happens to pipeline coverage if your paid CAC rises 15%? 25%? What's the break-even point where you'd reallocate budget to another channel?
The Uncomfortable Question
Some teams have been setting artificially high targets specifically to trigger overperformance. A $100 Target CPA that consistently delivers $60 looks like a win, and it was, until now. That arbitrage is closing.
The question worth asking in your next pipeline review: how much of our reported paid efficiency was real, and how much was the algorithm doing work we didn't pay for?
If the answer is "we don't know," you have five days to find out.
What This Means for Budget Conversations
Finance teams like predictability. This change, in theory, gives them more of it. A Target CPA of $50 should now mean a CPA closer to $50, not some variable number that depends on how much budget headroom the campaign has.
That's useful for planning. It's also useful for accountability. If you set a target and the system hits it, you own the number you chose. No more explaining why performance shifted when you added budget. No more variance between what you asked for and what you got.
The trade-off is that you lose the upside. The cheap conversions the algorithm was finding on its own? Those now require you to find them yourself, through better targeting, better creative, or lower targets that reflect what the market will actually bear.
The Five-Day Checklist
Before August 17:
- Pull all budget-limited campaigns using Target CPA or Target ROAS
- Calculate the gap between stated target and trailing 30-day actual
- Decide per campaign: lower target to lock in efficiency, increase budget to remove constraint, or accept higher CPA
- Update your CAC payback model with the new assumptions
- Brief your CFO or finance partner on the expected variance
This isn't a crisis. It's a correction. But corrections only stay small if you see them coming.
The teams that treat this as a media optimization task will fix their targets and move on. The teams that treat it as a forecast integrity issue will update their models, brief their finance partners, and avoid the Q3 conversation where someone asks why acquisition costs jumped 20% without warning.
Model or it didn't happen. You have five days.
