A Performance Max campaign still failing to convert at week six is not a Google problem. It is a signal problem the machine had to live with long enough to surface.
Every account we pick up with a failing PMax campaign has the same fingerprints: two or three conversion goals, asset groups sorted by product line, no audience signals worth mentioning, and a change history with edits every five days. The algorithm never stabilized. The CPA never landed. The operator concluded Performance Max does not work for their category.
Performance max not converting is a solvable problem in nearly every account we have seen - but only if you diagnose in the right order. This is the PMax Triage Stack: five layers to check in sequence before cutting budget or rebuilding from scratch.
Step 1: Establish whether the campaign has actually finished learning
The first question in any performance max not converting situation is whether the machine has had enough data to work from. The answer hinges on conversions, not days.
A PMax campaign needs roughly 50 conversions before the algorithm has a stable enough pattern to optimize against. The pmax learning period how long question is answered by conversion volume, not calendar. Budget set to generate two conversions per day puts you at 28 data points at the three-week mark. The machine is not underperforming - it is calibrating. CPAs will be erratic by design during this window.
- Check total conversions since launch, not days since launch. The pmax learning period how long answer is always "until you hit 50 conversions" - not two weeks, not four.
- If you are under 50, stop diagnosing and stop editing. Every material change partially resets the learning window and pushes the stable-data threshold further out.
- If you are under 50 and the change history already shows four or more edits, the window has been fractured multiple times. That is its own diagnosis before any other layer matters.
Step 2: Confirm you are optimizing for one clean goal
The most consistent cause of performance max not converting at the revenue level is a conversion goal setup problem. PMax optimizes toward whichever event fires most frequently - and most accounts have multiple events sitting as primary conversion actions without realizing it.
An account running "purchase" and "add to cart" as co-primary goals shifts toward add-to-carts within two weeks. The platform shows a healthy CPA. Revenue stays flat. The operator sees green metrics and a quiet checkout.
- Go to campaign settings and confirm exactly one primary conversion action is set at the campaign level.
- Every other event - page views, phone calls, email sign-ups, newsletter registrations - should be secondary or observed-only.
- If you cannot confirm this was set correctly before launch, assume it was not. Fix it now and expect the learning clock to reset. That is the right trade.
Step 3: Audit your asset groups for intent dilution
Performance max audience signals are the seeds the machine uses to find similar buyers. They only work if the asset group around them has a clear, consistent intent pointing in one direction.
When an asset group mixes multiple customer types - a new-season range and a clearance line in the same group, for example - the audience signals are pointing at different buyers and the machine cannot learn which pattern reliably predicts a conversion. The creative serves one message to the wrong person half the time.
- Each asset group should represent one customer intent with one offer. Not one product category - one reason a specific buyer converts.
- Add your highest-converting customer email list as performance max audience signals for each group. Not as a restriction on who sees ads - as a directional hint toward the signal pattern that matters.
- Add landing page URLs that already convert at a reasonable rate as URL expansion signals. This gives the machine a second reference point for what a real buyer looks like.
A DTC/ecommerce account separating "new collection buyers" from "clearance hunters" into distinct asset groups - with separate creative themes and separate audience signals - will consistently outperform the same account with both intents bundled together.
When a PMax campaign won't convert, the first instinct is to blame the machine. The data almost always points back to what the machine was asked to learn from.
Step 4: Open the Insights tab and read the search categories
Most operators check the Performance tab. The Insights tab is where PMax tells you how it is thinking - and it will tell you clearly when it has been pointed in the wrong direction.
Open the Insights tab and look at the search categories and audience segments driving spend. If the top spend categories are unrelated to your core buyer - broad shopping terms, informational queries, competitor brand categories - the asset group signals sent the machine somewhere it should not have gone.
Pull the top five spend categories and ask a simple question: would a buyer in this category be likely to convert on my specific offer? If the answer is no for two or more of the top five, the asset group needs a rebuild before any other fix will hold.
Step 5: Check for brand cannibalization with Search
Performance max not converting sometimes presents as underperformance when the actual issue is that PMax is winning auctions it never should have entered.
PMax will bid on branded search queries unless you have explicitly requested brand exclusions through Google support. An account without those exclusions uses PMax budget to win clicks from buyers who were already going to convert via Search - and those conversions appear in PMax, not in the Search campaign that generated the intent.
Check the Insights tab for your brand name in the search categories. If branded queries represent a significant share of PMax conversions, the campaign looks better than it is and your exact-match Search campaign is being starved of budget it should be capturing.
Submit the brand exclusion request through Google support. Then recalculate your true PMax CPA with branded conversions removed from the count - the number you see after that adjustment is the one worth acting on.
BAVai flags brand cannibalization every morning across the Google accounts we manage. It is a slow leak that compounds for weeks before it shows up in a manual audit, because it only becomes obvious when you compare PMax Insights categories against what branded Search is simultaneously reporting.
What the turnaround looks like
A campaign correctly diagnosed and rebuilt - one conversion goal, intent-based asset groups with clean performance max audience signals, brand exclusions in place, and a deliberate pause on edits until 50 conversions - typically stabilizes within three to four weeks of the fix. The CPA finds a floor. Volume grows from a real base.
Before that fix, every optimization attempt is building on an unstable foundation. The Signal-First PMax Setup gets this right from launch. The PMax Triage Stack gets you back to that foundation when the original setup did not.
The PMax Triage Stack checklist
- Check total conversions since launch - pause edits if under 50
- Review change history - count learning window resets
- Confirm exactly one primary conversion action at campaign level
- Rebuild asset groups around one customer intent each, not one product category
- Add highest-converting customer match list as audience signal per asset group
- Add high-converting landing page URLs as URL expansion signals
- Open Insights tab - check top five spend categories against your core buyer profile
- Submit brand exclusion request through Google support if not already done
- Recalculate PMax CPA with branded conversions removed
- Set the next review date at 50 conversions - not a calendar date
When you look at your PMax campaign's change history, does it tell a story of clean signals and patient learning - or a story of edits chasing erratic numbers that were never stable enough to read?