If Google Ads is reporting a 4x ROAS and a CPA under target, you have every reason to feel good. The dashboard looks clean. The report goes out. Everyone is comfortable.
Then someone asks the uncomfortable question at the end of the quarter: "Why aren't we growing?" Revenue is flat. Profit is the same. And between the numbers in Google Ads and the numbers in accounting, there's a gap nobody can explain.
A recent article on Search Engine Journal nails it: Smart Bidding does exactly what you tell it. Not what you need. And that's a problem we see regularly in PPC audits.
Smart Bidding Does What You Tell It. Not What You Need.
Google says it clearly: Smart Bidding optimizes for the target you set. If you told it to deliver 4x ROAS, that's what it will do. It doesn't ask whether that ROAS translates into profit, whether the customers it brings stick around, or whether the business is actually growing.
Think of it like GPS. It takes you exactly where you asked. The problem shows up when the destination you entered six months ago is no longer the right one. Costs changed. Margins shifted. Product mix evolved. But the target stayed the same.
A 4x ROAS at a 60% margin is one thing. A 4x ROAS at a 20% margin is a completely different story. The algorithm doesn't make that distinction. You have to.
On top of that, a Google bidding update from August 2026 pulls budget-constrained campaigns closer to their set target, instead of letting them outperform it. This means an outdated target doesn't just stagnate performance. It becomes actively costly.
And here's the uncomfortable part: because Smart Bidding does a solid job on the metric you gave it, nobody looks further. The report shows green. Nobody asks what's behind the green. The difference between an account that performs and an account that grows only becomes visible when you put the ad numbers next to the ERP or CRM data.
Where the Connection Between Ads and Profit Breaks Down
From our agency experience, the problems come from a few predictable directions.
Existing customers dominate conversions. Smart Bidding naturally gravitates toward people who already know you. Their conversion is cheap, ROAS looks excellent. But the new customer base isn't growing. You're essentially paying Google to bring back people who would have come anyway.
We saw this firsthand with an e-commerce client: a consistent 5x ROAS for six months, but flat revenue. When we dug in, over 65% of conversions came from returning customers. The algorithm had found the path of least resistance and wasn't leaving it.
Brand terms inflate your numbers. Brand search campaigns show spectacular ROAS. But think about it: if someone searches your exact brand name, how much credit does the ad deserve? Much of that ROAS comes from demand that already existed. Without incrementality testing, you can't tell how much is real and how much is artificially attributed.
Proxy conversions are not sales. A filled form, a downloaded PDF, a booked demo. Smart Bidding treats them as success. But if you don't have a feedback loop with your CRM that tells the system which leads actually became sales, you're optimizing for quantity, not quality. The algorithm will bring cheap leads, not good ones.
The problem gets worse when you have multiple conversions set as primary. Smart Bidding treats them equally: a newsletter signup and a quote request carry the same weight. If you don't differentiate, the algorithm will naturally gravitate toward the easiest conversion to get, not the one that matters.
Tight efficiency caps reach. When the target is rigid, the algorithm skips more expensive auctions that would have been profitable. You lose impression share with new audiences. The account is "efficient," but the addressable market stays unexplored. We wrote about preparing your ad account for peak periods, where this matters even more.
How to Spot the Disconnect in Your Account
Before any optimization, ask yourself a few concrete questions.
When did you last recalculate your ROAS or CPA target based on current margins? If the answer is "at setup," you have a problem. Margins move. The target should move with them.
What percentage of conversions comes from new vs. returning customers? If you don't have this report segmented, build it. It's the first check we run in any audit. If returning customers make up more than 50%, the algorithm has optimized for comfort, not growth.
How much of your ROAS comes from brand campaigns? Isolate brand search from generic. Compare. If brand contributes 60%+ to overall ROAS, the numbers are hiding more than they show.
Is the primary conversion in the account a real sale or a proxy? If it's a proxy (lead, form, booking), do you have a closed-won feedback system? Without it, Smart Bidding can't distinguish a good lead from one that will never buy.
Are you losing impression share because of the target? If so, test with a more relaxed target over a limited period. Sometimes a CPA that's 15% higher brings significantly more volume, with better net profit.
What to Do After You Find the Gap
The first step isn't to change the target. It's to understand why the current target no longer works. Recalculate based on real margins from the last quarter. If margins shifted by more than 5 percentage points, the ROAS target needs updating.
Separate brand from non-brand reporting. Set a different ROAS target for each. Brand will always have a higher ROAS, but it shouldn't dictate the overall account target.
If you're running proxy conversions, implement offline conversion import from your CRM. Google Ads can receive data about which leads turned into actual sales and optimize accordingly. It's one of the highest-impact changes you can make, and one that most companies keep postponing.
And perhaps most importantly, check your new vs. returning customer split monthly. Not quarterly. Monthly. If the share of new customers drops two months in a row, that's a clear signal the algorithm has gone on autopilot.
Your Google Ads numbers don't lie. But they don't tell the full story either. An account that hits every target can be an account that simply stopped growing. And that's not the algorithm's fault. It's the assumption that the target from six months ago still reflects today's reality.
Check. Segment. Test. And above all, put the ad account data in the same room as the business data. That's where the real answers are.
