On this page
Does Performance Max hit its target CPA?
It depends on the sample, and our two samples disagree. In the original one, Performance Max overshot its target far more often than Search. In a newer, smaller one from 2026, it overshot less often.
The same question applies to any campaign bidding to a target, and our other research on bidding and budgets asks it more widely.
Target CPA bidding means you name the cost per sale you are willing to pay and the system bids to hit that number on average. Google says plainly that some conversions will cost more and some less. The question nobody answers with data is how much more, and for whom.
Hand that instruction to a Search campaign in the original sample and it came in more than a tenth above target in 26.3% of cases. Hand it to Performance Max and it did so in 71.3%.
The medians said it from the other side. Half of the Performance Max campaigns delivered at 1.30x their target or worse, against 0.95x for Search. The typical Performance Max campaign in that sample was not missing narrowly. It was running about 30% above the number it was set, and 21.3% of them delivered at twice the target or worse.
Two honest limits before you carry that number anywhere. It rests on 178 Performance Max campaigns, not on the full 2,723, and Performance Max was not present in every account in this portfolio. Nothing here says the campaign type causes the overshoot: advertisers choose where to run Performance Max, and those are often the goals Search already struggles with.
Update, October 2026: a newer sample reverses the comparison
A newer, smaller sample from the managed portfolio points the other way. Between January and May 2026, 27.7% of Performance Max campaigns with an explicit target came in more than 10% above it, against 45.5% of Search campaigns. In the original sample those shares were 71.3% and 26.3%. The median Performance Max campaign delivered at 0.94x its target and the median Search campaign at 1.07x, so the original miss rate is not a fixed property of Performance Max.
The obvious objection is that the two campaign types sit in different accounts, so the totals compare accounts as much as types. That is why we also compared the two types inside the same accounts. For every account that ran both, we took the share of its Performance Max campaigns more than 10% above target and subtracted the same share for its Search campaigns. The median gap was 12.9 percentage points in Performance Max’s favour, so the reversal shows up inside shared accounts, not only in the totals.
The two samples differ in size, months and campaigns, so neither shows the product getting better or worse over time. What they do show is that the answer to this article’s opening question changes with the campaigns you measure. For your account, the honest starting point is your own numbers, which is where the steps later in this article begin.
How to check your own Performance Max target
Compare Performance Max with Search inside the same account and reporting window. For each campaign, divide delivered cost per conversion by its target, then compare the share more than 10% above target in each type.
Do not reuse your Search target as a Performance Max success bar. Median Performance Max delivery was 1.30x target in the original sample and 0.94x in the newer one. Neither is a safe allowance for your account. Check profit and total conversions alongside target adherence.
For the broader question, use how target CPA performs across the portfolio. This comparison isolates Performance Max against Search; it does not establish a miss rate for every bidding strategy.
If you need that comparison across reporting windows, a Profit Forensics examination can inspect your own campaigns.
When this does not apply
Three conditions make these numbers the wrong lens for your account.
- You run almost no campaigns with an explicit target. If your account sits on Maximise Conversions or manual bidding, there is no promise to keep and nothing here to measure.
- Your campaigns are small and few. Under roughly ten conversions per campaign per window, the ratio swings on noise rather than on the bidding system, which is why that was the inclusion floor here.
- Your conversion actions changed inside the window. A new conversion definition, a switch to value-based goals or a tracking fix moves the delivered cost per conversion for reasons that have nothing to do with the target.
The performance periods measured here end before Google’s 17 August 2026 change, under which eligible budget-limited Target CPA and Target ROAS campaigns are steered towards their target instead of beating it. For an eligible budget-limited campaign, read them as the baseline before that change.
Researcher’s take
I would not budget for a fixed Performance Max overshoot. The original comparison makes it look much worse than Search; the newer one reverses that order, including inside shared accounts. My next step would be the same in either case: compare the two types in the account I am deciding about, using the same dates and conversion definition. A portfolio comparison tells me what to inspect. It does not give me a safe mark-up to add to a client’s target.
Igor Ivitskiy, PhD, Doctor Ads
Key takeaways
- Original overshoot: 71.3% for Performance Max against 26.3% for Search.
- January to May 2026 overshoot: 27.7% for Performance Max against 45.5% for Search.
- Median Performance Max delivery changed from 1.30x target to 0.94x across the separate samples.
- The newer paired-account comparison favoured Performance Max by 12.9 percentage points. Measure your own comparison before judging the campaign type.
Frequently asked questions
Does Performance Max miss its target CPA more often than Search?
It depends on the sample, and this study measures the miss, not its cause. In the original sample, 71.3% of 178 Performance Max campaigns delivered more than 10% above target and the median landed at 1.30x, while 1,563 Search campaigns in the same accounts sat at 26.3% and 0.95x. In a newer sample from January to May 2026, 65 Performance Max campaigns came in at 27.7% and 0.94x, and 145 Search campaigns at 45.5% and 1.07x. Advertisers also choose what to put into each campaign type, so part of any gap belongs to that choice rather than the product.
If Performance Max overshoots its target, is it worth using at all?
That is a profit question, not an adherence question, and this data deliberately does not answer it. A campaign type that runs 30% above its target can still be the most profitable line in the account if it brings conversions the other campaigns never reached. What the numbers do say is that target adherence is a shaky way to judge it. About seven Performance Max campaigns in ten ran more than 10% over target in the original sample, and fewer than three in ten in the newer one. Judge it on profit and total conversions instead.
Method and sources
Doctor Ads Profit Forensics is the research desk of a Google Ads practice with 19 years and a large managed spend sample behind it.
Figures come from a forensic analysis of a managed portfolio of Google Ads accounts, measured over each account’s primary reporting window, most spanning September 2024 to February 2025. The population for this study is every campaign carrying an explicit target CPA with at least 10 conversions and non-zero cost, deduplicated to one row per campaign: 2,723 campaigns across the accounts in our managed portfolio and 7 verticals. Accuracy is the campaign’s own delivered cost per conversion divided by its own target, both in the account’s own currency, so the ratio is unitless while cost levels are never pooled across accounts. Multiples are rounded to two decimals throughout. The denominator is the campaign’s configured target as exported, not Google’s traffic-weighted average target, so mid-window target edits and device adjustments can inflate an apparent miss; recomputing on average target is the natural follow-up study. Campaign type was not randomly assigned and this portfolio carries client self-selection, so the comparison is descriptive rather than causal. Related reading: portfolio bid strategies and the dead hours in your account.
The October 2026 update uses January to May 2026 performance from a 10 July 2026 snapshot of the managed portfolio, with one account refreshed in September 2026. For each campaign, cost and recorded conversions are summed in the account’s own currency, and the resulting cost per conversion is divided by the target CPA configured in that snapshot. The original rules apply: a positive target, positive cost and at least 10 conversions, with overshoot meaning a delivered cost per conversion more than 1.10 times the target. In total, 223 campaigns of all types qualify; 65 of them are Performance Max and 145 are Search. Medians weight each campaign equally. The paired check covers the accounts that ran qualifying campaigns of both types and takes the median of each account’s Performance Max overshoot share minus its Search overshoot share. June is left out so that recent conversions have time to arrive. Targets come from one snapshot, so a target edited during the window can make a miss look larger or smaller than it was, and advertisers still choose which goals go to which campaign type. The update measures recorded target adherence, not incremental sales, profit or a change in the product over time.