Performance Max spending the daily budget is not, by itself, evidence of waste. The budget is permission to spend, and automated bidding is built to enter auctions when it predicts the selected goal can be achieved. The real diagnosis is whether the extra spend produced acceptable marginal business value.
Work in this order: contain unacceptable risk, identify what changed, verify the conversion signal, locate delivery, then change the smallest input supported by evidence. Stories about "the algorithm" come last.
Contain unacceptable risk first
If purchase tracking is duplicated, lead spam is being counted, a promotion has ended or spend threatens cash flow, reduce or pause the campaign. Preserving learning is not more important than preventing known loss. Record the exact time and reason.
If the campaign is merely below target, set a temporary budget that caps downside while you investigate. Avoid simultaneous changes to budget, target, products and creative because the result will be impossible to interpret.
1. Confirm the pattern
Compare recent complete days with a preceding comparable period and the same weekday mix where relevant. Add Cost, Conversions, Conversion value, Value/cost, Cost/conv., Budget and Bid strategy type. Exclude today's partial data.
Check absolute PMax spend and its share of total account spend. A larger share caused by another campaign being paused is different from a PMax surge. Review total account revenue and contribution, not only the campaign row.
2. Reconstruct what changed
Open Campaigns > Change history and filter to the campaign. Look for budget and target edits, new primary conversions, location changes, product filters, customer-acquisition settings, assets, search themes, Final URL expansion and applied recommendations. Add feed, price, stock, promotion and site changes from outside Google Ads.
A lower target ROAS usually permits broader bidding; a new primary micro-conversion can redirect optimisation. Treat timing as evidence, but do not assume correlation proves the cause.
3. Find the conversion signal that justified spend
Select Segment > Conversions > Conversion action. Separate purchases or qualified leads from page views, calls, form starts and other softer actions. Then segment by Ad event type where available to distinguish click, engaged-view and impression-associated conversions.
If a cheap secondary outcome became Primary and grew while sales did not, the campaign may be following the instruction precisely. Correct goal inclusion, values and counting before blaming traffic. The Performance Max setup checklist covers the pre-launch measurement work.
4. Locate spend by channel and format
Go to Insights & reports > Channel performance. Add Cost, Conversions, Conversion value and Value/cost. Use the channel dropdown, segment by Ads using product data or Ads using video, and review the time-series chart. The Status column can reveal missing assets, product issues and limited channel eligibility.
Google's channel performance guide documents Search, Display, YouTube, Discover, Maps, Gmail and Search partners, with report data from 6 June 2025 onwards. This can establish that YouTube spend rose after a video was added. It cannot prove that forcing the same money into Shopping would have produced more profit.
PMax does not provide channel-level budgets. Content suitability and placement exclusions address brand safety; using them only to manufacture a preferred mix can reduce reach without proving better economics.
5. Inspect search terms and brand dependence
Open Search terms, sort by cost and add conversion value. Classify terms as own brand, product, generic category, competitor, irrelevant or ambiguous. Google's PMax search-terms documentation confirms granular search terms and exclusions are available.
Brand traffic is a measurement choice, not automatically theft. If the campaign's job is non-brand acquisition, apply a brand exclusion and expect the reported ROAS to change. Google says PMax brand exclusions apply to Search, Shopping and YouTube Search inventory, with a separate option for Shopping ads. They are not a universal cross-channel holdout.
Use negative keywords for clearly irrelevant or unsafe terms. Google recommends brand exclusions for brands because they cover variants more comprehensively.
6. Find products absorbing the money
For retail, open Campaigns > Products. Add item ID, product type, custom labels, cost, conversions, value/cost and status. Look for low-margin items with flattering revenue ROAS, new products entering Everything else, unavailable bestsellers shifting spend elsewhere and products overlapping other campaigns.
Product reporting expanded in June 2026 across all PMax networks. It still does not contain your margin unless you provide or join it. Use the product feed optimisation guide when eligibility, attributes or grouping are the underlying issue.
7. Audit landing pages and expansion
Review Landing pages by cost. Check whether traffic reached products or categories that can complete the advertised promise. Final URL expansion is on by default and may replace the supplied URL with another page on the domain.
Exclude careers, policy, support, account and irrelevant editorial URLs. Turn expansion off when one destination is mandatory. Google's Final URL expansion guide notes that a page feed does not become a strict allow-list while expansion remains enabled.
8. Test the target against business economics
Budget controls exposure; the bid strategy controls the outcome pursued. If the target is looser than the commercial floor, PMax can rationally buy conversions that lose contribution. If the target is unrealistically strict, delivery may become constrained or volatile.
Hypothetical example: a retailer spends £300 per day at 400% revenue ROAS but needs 500% after product cost, fulfilment and returns. The campaign is not "profitable but spending too much." Its optimisation threshold is below the business requirement. Lowering budget limits loss; correcting value inputs, product scope or target addresses the rule.
Account for conversion lag and avoid changing a target because one late high-value order moved the average.
9. Check structure and assets
Use the asset-group table to compare propositions, then inspect asset details and combinations. Missing formats can limit channel eligibility; weak claims can waste attention. But a Low asset label does not explain an account-level spike by itself.
If a range needs a different target or protected budget, move it to a separate campaign. Another asset group does not create either control. The asset-group structure guide provides the decision test.
Match the action to the evidence
| Observed issue | Primary action | Do not claim |
|---|---|---|
| Wrong conversion grew | Correct goals and values | Audience quality caused it |
| Irrelevant queries spent | Add precise negatives; review themes and pages | All Search delivery is poor |
| Brand dominates value | Set a documented brand policy | Every brand sale is incremental |
| Low-margin items spend | Fix values, labels, scope or targets | Revenue ROAS equals profit |
| Unexpected pages receive traffic | Use URL controls | The entered final URL was a hard limit |
| Channel mix changed | Check assets, goals, targets and timing | The channel change caused the result |
Review the intervention
Annotate the change and wait long enough to include normal conversion lag and enough business outcomes for a decision. Reconcile Google Ads results with backend revenue, contribution, new customers, refunds and lead quality.
If PMax ROAS rises while total account contribution falls, the ratio is not a scaling case. If reported ROAS falls after brand exclusion while new-customer contribution improves, the stricter measurement may be doing its job.
What reporting cannot prove
Search terms, channels, products and landing pages explain observed delivery. Standard attribution cannot show how many conversions would have occurred without PMax. Cannibalisation and incrementality remain inferences until tested.
Use a supported PMax experiment, geographic test or customer holdout when the budget decision warrants causal evidence. Until then, state which facts are observed and which conclusion is inferred.