A Performance Max campaign that isn't converting rarely has an algorithm fault. ~90% of cases trace to feed quality, asset-group structure, or brand exclusions.
What "PMax not converting" actually means
"Not converting" is a symptom, not a diagnosis. Separate four modes first: zero conversions ever (tracking/eligibility), CPA above target (bidding or signal), drop after stable period (fatigue, feed disapproval, competitor), or Ads-vs-CRM mismatch (attribution). Each has a different fix. PMax also looks "broken" in its first ~2 weeks while Smart Bidding learns (Smarter eCommerce, 2025).
Top causes
Ranked by frequency in our 2025-2026 audits of Performance Max accounts:
- Conversion tracking misconfigured. Goals not attached to the campaign, duplicate conversion actions inflating counts, or Enhanced Conversions silently failing — Smart Bidding optimises towards the wrong signal.
- Feed quality below threshold. For Shopping-led PMax, item-level disapprovals, missing GTINs, low-resolution images, or thin titles cap eligible inventory. Optmyzr's 2025 PMax benchmark report flags feed quality below ~80% as the dominant predictor of underperformance (Optmyzr 2025 PMax Benchmarks).
- Over-segmentation. Five PMax campaigns each starved of data convert worse than one consolidated campaign. Smart Bidding needs ≥30 conversions per month per campaign to stabilise (Smarter eCommerce, 2025).
- Missing brand exclusions. Without them, PMax cannibalises branded search at lower CPCs and inflates reported conversions that would have happened organically.
- Asset-group themes too broad. One asset group covering five product categories starves Smart Bidding of theme-level signal.
- Audience signals treated as targeting. Signals only suggest — they are not segments. Loading 20 audiences as "observation" produces noise, not lift.
- Insufficient learning runway. Judging before ~2 weeks or after a >20% budget swing resets learning (Search Engine Land, 2025-06).
Diagnostic checklist
Run these in order. Stop at the first failure — fixing it usually moves the campaign without further changes.
| # | Check | Where | Pass criteria |
|---|---|---|---|
| 1 | Conversion goals attached | Campaign settings → Conversion goals | Account-default OR explicit primary action set; no orphaned goals |
| 2 | Conversions imported in last 7 days | Tools → Conversions | Status = Recording, no "No recent conversions" warning |
| 3 | Feed quality (Shopping) | Merchant Center → Diagnostics | ≥80% items active, no critical disapprovals |
| 4 | Channel split | PMax campaign → Insights → channel-level performance | Conversions distributed across ≥2 channels; no >70% spend on Display alone |
| 5 | Search terms report | PMax campaign → Insights → search terms | ≥60% commercial-intent queries; brand queries excluded if exclusions are on |
| 6 | Brand exclusions | Campaign settings → Brand exclusions | Configured if branded search is a meaningful revenue share |
| 7 | Asset group count | Campaign view | ≤1 group per coherent product theme; consolidated when conversions <30/mo |
| 8 | Learning phase status | Campaign status column | "Eligible" — not "Learning" — for at least 7 days |
Google Ads Help describes the gate plainly:
"Performance Max campaigns use machine learning to optimize bids and placements to drive conversions or conversion value for your goals."
— Google Ads Help: About Performance Max, retrieved 2026-04-29
The implication: starve the model of conversion signal and "machine learning" becomes "expensive guessing".
Pattern from our audits: three fixes resolve most "PMax not converting" tickets without touching bidding. First, consolidate — accounts running 4-6 PMax campaigns at <10 conv each recover after merging into 1-2 that cross the 30-conv/month gate. Second, exclude the brand — much "great PMax performance" is branded search rebought at a 30% premium. Third, screenshot the channel split at launch and at week 2; if Display+Gmail climbs above 60% while Shopping/Search falls, audience signals or assets are pulling budget into low-intent inventory. None require pausing.
"How long should I wait before judging PMax?"
At least 2 weeks of stable budget and unchanged goals — the learning-phase exit window for tCPA/tROAS (Search Engine Land, 2025-06). Budget swings >20%, new conversion actions, or asset-group restructures restart the clock. 14 days in with stable inputs and no conversions = structural issue, not learning.
"Should I pause PMax if it's burning budget with no conversions?"
Pause only if rows 1-2 surface a tracking failure. Otherwise drop daily budget 50% — pausing resets learning and forfeits ~2 weeks of signal. Investigate channel split and search terms first.
"Why does PMax look profitable but my CRM shows fewer real leads?"
Three causes: duplicate conversion actions, raw form-fills counted before qualification, or branded-search cannibalisation. Fix: deduplicate, switch primary conversion to a qualified-lead event via offline import, configure brand exclusions.
Fix paths
- Tracking broken: start with conversion tracking and verify Enhanced Conversions are firing.
- Over-segmented or starved: review the Performance Max overview for the 30-conv/month gate, then consolidate.
- Asset groups misshaped: see asset groups — one theme per group, audience signals as suggestions only.
- Bidding unstable: check Smart Bidding for tCPA / tROAS prerequisites and learning-phase rules.
- Branded cannibalisation suspected: configure brand exclusions and review negative keywords.
When to escalate
Escalate to a human reviewer when all 8 rows pass but the campaign still under-delivers after 21 days of stable budget. Causes are usually outside campaign settings: landing-page collapse, broken consent banner, supply-side change, account-level policy review. Do not pause first — pull a 30-day search-terms report, channel-split screenshot, conversion-action audit, and Merchant Center diagnostics export, then escalate. Pausing destroys the signal a reviewer needs.
Sources
- Google Ads Help: About Performance Max — official campaign-type definition + Smart Bidding framing, retrieved 2026-04-29
- Smarter eCommerce: State of Performance Max 2025 — 30-conv/month threshold + over-segmentation evidence
- Optmyzr 2025 PMax Benchmarks — feed-quality 80% threshold, channel-split distributions
- Search Engine Land: Performance Max learning phase guide (2025-06) — 2-week learning window, budget-change reset rules
- Smarter eCommerce: PMax 2025 FAQ — channel-level + asset-level reporting walkthrough, retrieved 2026-04-29