Short version: A Performance Max campaign that spends without converting almost always has one of seven causes, and the order you check them in matters more than the checks themselves. Start with conversion tracking, because a campaign fed broken conversion data is optimizing toward nothing and every other change is noise. Then separate real new-customer revenue from brand harvesting, audit the product feed, judge the learning period on conversions rather than days, and only then look at structure, targets, and budget. Change one thing at a time, give each change room to settle, and accept that for some businesses Performance Max is simply the wrong tool. Rebuilding the campaign is the last resort, not the first move.
The search that brings people to this page usually happens somewhere around day ten. The campaign launched, the spend arrived on schedule, and the conversions did not. Google's interface offers no explanation, the recommendations tab suggests raising the budget, and every article you find says either "wait longer" or "PMax is a scam". Neither is a diagnosis. What follows is the order we work through when an ecommerce Performance Max campaign is spending but not converting, and it is an order for a reason: each check either finds the cause or rules out a category of causes, so you never fix the wrong thing first.
Why Performance Max fails quietly
Performance Max is a black box by design. You hand Google a feed, assets, a conversion goal, and a target, and it decides which queries, placements, audiences, and formats to buy. When that works, it works with less effort than anything else in the platform. When it does not, the same opacity that saved you effort now hides the cause. The campaign will not tell you that your conversion tag double-fires, that your titles lose every Shopping auction, or that your target is a number the account has never achieved. It just spends, because spending is what it is built to do.
The black box cuts both ways in a second sense too. A PMax campaign that looks brilliant can be quietly harvesting conversions that would have happened anyway, and a PMax campaign that looks broken can be honestly reporting that your product economics do not support cold acquisition. The reported number and the real performance can diverge in either direction, which is why the diagnosis below starts with measurement rather than with the campaign.
The diagnostic order
1. Verify conversion tracking before touching the campaign
This is first and it is not optional. Performance Max optimizes toward the conversion data you feed it, so if that data is wrong, the campaign is executing a broken instruction perfectly. Check that the primary conversion action is the purchase, not add to cart or page view. Check for duplicate tags firing two conversions per order, which inflates reported results while training the bidding on false signals. Check that conversion values are passing correctly, in the right currency, because a value-based strategy with missing values has nothing to bid on. If the store runs on Shopify, our guide to Shopify conversion tracking covers the specific failure points, and the full diagnostic for numbers that look wrong account-wide is in the conversion tracking troubleshooting guide. Until tracking is verified, every other observation about the campaign is unreliable, so resist the urge to skip ahead.
2. Ask whether it was ever really converting
The second check runs in both directions. If the campaign used to convert and stopped, look at what those conversions actually were. Performance Max loves branded searches and existing customers, because they convert cheaply and make the campaign look effective. A PMax campaign whose reported revenue is mostly people who searched your brand name is not driving growth, it is billing you for it. Conversely, if conversions collapsed right after you added brand exclusions or customer-acquisition settings, the campaign may now be doing honest work for the first time, and the honest number is simply worse than the flattering one was. That is not a malfunction. It is the real cost of new-customer acquisition showing itself, and the fix is economics and creative, not settings. Segment your results by new versus returning customers and by branded traffic before concluding anything else.
3. Audit the feed, because Shopping placements decide most of it
For ecommerce accounts, the majority of Performance Max conversions typically come through Shopping-style placements, and those are won or lost on the product feed. Titles that lead with the brand name instead of what the product is, images that blur at thumbnail size, and prices that sit well above the competing listings will lose auctions no matter what the campaign settings say. Pull up a few of your products in the Shopping results a customer would actually see and compare honestly. If your listing would not win your own click, the campaign cannot save it. This is usually the highest-leverage fix on the whole list, and we cover it in depth in the product feed guide.
4. Judge the learning period on conversions, not days
Performance Max needs conversion volume to calibrate, and judging it after five days and a handful of conversions is judging noise. A reasonable standard is to wait for roughly thirty conversions or four to six weeks, whichever comes first, before drawing structural conclusions. But the learning period excuse has a boundary. "It is still learning" is not an answer after two months of spend with no trend toward target, and an algorithm cannot learn its way out of broken tracking, an uncompetitive feed, or an impossible target. Use the learning period to withhold verdicts, not to withhold diagnosis. The first three checks on this list are valid from day one.
5. Look at the structure
Two structural patterns produce the spending-without-converting symptom. The first is too many asset groups or campaigns splitting a modest budget into slices too thin for any of them to gather signal, so everything stays in learning forever. The second is the opposite: one asset group mixing bestsellers with dead stock, where the algorithm spends exploration budget on products that have never sold and never will. The right shape for most stores is a small number of asset groups organized around genuinely different product economics or audiences, each with enough budget to learn. If you have seven asset groups and a budget that supports two, the structure is the problem.
6. Check the target against what the account has actually achieved
A target ROAS is an instruction, not an aspiration. Set it far above anything the account has historically achieved and Performance Max responds in one of two ways: it throttles spend because almost no auction clears the bar, or it chases whatever junk inventory technically pencils out at that number. Look at the campaign's actual trailing ROAS, sanity check it against your real margins, and set the target near what the account has demonstrated, then move in small steps. An aspirational target does not motivate the algorithm. It starves it.
7. Check whether the budget can exit learning at all
Last on the list because it is only meaningful once everything above is clean. If your average order requires, say, forty clicks to produce a sale, a daily budget that buys six clicks will take weeks to accumulate the conversions the system needs, and the campaign will sit in a permanent learning state that looks exactly like failure. Work backwards from your conversion rate and average CPC to the daily conversion count your budget can support. If the answer is less than roughly one conversion a day, either concentrate the budget into fewer campaigns or accept a much longer evaluation window.