Performance Max can look efficient while answering the wrong business question. A campaign may report an acceptable return because it captures branded demand, emphasizes returning customers, optimizes toward shallow leads, or concentrates spend in inventory that would not have received the same budget under a channel-level plan. The reverse is also possible: a useful campaign can look weak because conversion values are incomplete, offline outcomes arrive late, or the account evaluates a prospecting campaign against a blended target it was never designed to meet.
A useful Performance Max audit therefore does not start with the recommendation to add assets, change a target, or pause the campaign. It starts by identifying what the campaign is supposed to accomplish and whether the measurement system can observe that outcome.
The short answer: audit Performance Max in four layers—business outcome, measurement, delivery, and incrementality. Then classify the campaign as keep, fix, constrain, or replace. Keep it only when its role is explicit and its economics survive reconciliation outside the Google Ads interface. Fix it when the campaign is directionally appropriate but its goals, feed, assets, or data are defective. Constrain it when valuable automation needs tighter brand, URL, query, or inventory boundaries. Replace it when the campaign format conflicts with the buying motion or fails a credible incremental test.
This guide is designed for growth leaders and paid-media operators evaluating an existing account. It is not a promise that a specific setting will improve ROAS. The output should be a decision record: what the campaign is for, what the evidence says, which failure mode is present, and which change can be tested without destroying the baseline.
What a Performance Max audit should answer
Before opening the settings panel, write down five answers:
What business outcome is this campaign buying? Examples include first orders from new customers, qualified pipeline, store visits, or profitable repeat purchases.
What role does Performance Max play in the account? It may be an incremental cross-channel layer, the primary shopping engine, a lead-generation campaign, or a catch-all campaign created without a clear role.
Which conversions and values influence bidding? The campaign can only optimize toward the conversion actions and values made available to it.
What demand is the campaign allowed to capture? Brand, non-brand, existing customers, new customers, specific product groups, and geographic markets should be intentional choices.
What evidence would justify keeping, constraining, or replacing it? Define the decision rule before reviewing a favorable dashboard.
Google describes Performance Max as a goal-based campaign type that can access inventory across Search, YouTube, Display, Discover, Gmail, and Maps. It uses the advertiser's conversion goals, budget, audience signals, creative assets, and optional data feeds as inputs (Google Ads Help: About Performance Max campaigns). That breadth makes the campaign useful, but it also means that an account-level result can hide materially different jobs. An audit must reconnect the aggregate result to the business job.
The four-decision framework
Decision Use it when Primary action Evidence required ------------ Keep Goals, values, customer definitions, and tracking are credible; delivery fits the campaign's role; economics reconcile Preserve the baseline and document monitoring thresholds Stable business outcome, explainable channel mix, no material measurement defect Fix The campaign type may fit, but inputs or measurement are broken Repair one causal layer before judging the format Identified defect in goals, values, feed, assets, URLs, or offline feedback Constrain Automation finds value, but captures demand or inventory outside the intended role Apply the narrowest effective control and test the result Evidence of brand leakage, unsuitable URLs, irrelevant queries, inventory risk, or customer-mix mismatch Replace The required control, transparency, or buying motion is incompatible with the campaign, or incrementality is weak Shift budget through a controlled experiment, not an abrupt opinion-led shutdown A credible alternative, pre-defined success metric, and sufficient observation window
“Kill it” is rarely a useful audit category. It collapses several different diagnoses into one action. If tracking is wrong, pausing the campaign does not repair the evidence. If branded demand is the issue, a brand exclusion may be more informative than shutting off all inventory. If the business needs exact query control, standard Search may be the better tool for that job, but Performance Max could still have a separate role.
Layer 1: audit the business outcome
Map the optimization event to an economic event
List every conversion action included in the campaign's “Conversions” column and classify it:
revenue-bearing purchase;
qualified or converted lead;
early funnel action such as form submission, account creation, or trial start;
engagement proxy;
imported offline outcome;
duplicate or legacy event.
The audit question is not merely whether a tag fires. It is whether bidding receives an outcome that is close enough to commercial value. A lead-generation campaign optimized to all form submissions can favor low-friction, low-quality demand. An ecommerce campaign with flat or missing values can treat unlike orders as equivalent. Google's own guidance recommends choosing conversion goals that matter to the business and, for lead generation, optimizing toward qualified or converted leads as close to the final sale as practical (Google Ads Help: Multiply conversions with Performance Max).
Create a simple conversion map:
Conversion action Primary or secondary Value source Delay Business owner Decision use ------------------ Purchase Primary Transaction value Same session or delayed Ecommerce Bidding and revenue reconciliation Qualified lead Primary when reliable CRM stage or expected value Days or weeks Sales Bidding and pipeline quality Form submitted Secondary when downstream data exists None or provisional Immediate Marketing Diagnostic only Page view / engagement Secondary None Immediate Marketing Never a revenue substitute
Do not assign monetary values merely to make value-based bidding available. A fabricated value hierarchy gives the system precise-looking but false instructions. If downstream values are not ready, document that limitation and fix the data path before interpreting reported value as business value.
Define the customer the campaign should acquire
For ecommerce, distinguish total revenue from new-customer revenue. For subscription or lead-generation businesses, distinguish account creation from activation, qualification, and payback. Google's customer lifecycle goals can prioritize new customers, new high-value customers, or re-engagement, but their usefulness depends on reliable first-party customer definitions and compatible conversion goals (Google Ads Help: About customer lifecycle goals).
Audit the underlying customer lists, tagging logic, and value assumptions before treating a “new customer” label as ground truth. Unknown customers and returning buyers who use a different identifier can change the reported mix. The setting is an optimization input, not an independent incrementality study.
For a broader budgeting view, connect this audit to the distinction between MER and platform ROAS. Platform reporting can inform campaign operations; blended economics and contribution margin determine whether the acquisition program is sustainable.
Layer 2: audit measurement before media
Reconcile three ledgers
A Performance Max audit should compare at least three ledgers over the same time boundary:
Google Ads: conversions, conversion value, cost, and attribution settings.
Analytics or first-party event data: sessions, orders, qualified events, and campaign identifiers.
Finance, commerce, or CRM system: net revenue, cancellations, returns, qualified pipeline, closed revenue, or contribution margin.
The totals will not match perfectly because the systems use different attribution rules, identity resolution, time zones, and event timestamps. The goal is not forced equality. The goal is to explain the major differences and detect discontinuities. A sudden improvement in Google Ads with no corresponding movement in orders, qualified pipeline, or blended economics is an audit signal, not proof of failure or fraud.
Use a reconciliation table:
Check Question Failure signal Next action ------------ Event count Are primary conversions duplicated or missing? Step change without a business change Inspect tags, imports, deduplication, and consent behavior Value Does reported value reflect net commercial value? Gross value rises while margin or net revenue falls Separate diagnostic revenue from budgeting economics Delay How long until the outcome is observable? Recent periods are judged before outcomes mature Use mature cohorts or a delay-adjusted window Identity Can customer and lead outcomes be reconciled? Large “unknown” segment or missing offline stages Repair first-party matching and CRM import workflow Attribution What event receives credit and when? Channel result changes after model or window change Annotate the change and preserve a comparison baseline
If the event pipeline is the problem, the appropriate next step is a measurement repair, not a bidding change. The server-side tracking guide explains how to build a conversion system that can be reconciled without treating a server event as automatic proof of accuracy.
Check goal scope and account defaults
Review whether the campaign uses account-default goals or campaign-specific goals, and whether legacy actions remain primary. Confirm that duplicate GA4 and Google Ads conversions are not both steering bidding for the same business event. Verify that test transactions, internal leads, and low-value actions are excluded or secondary where appropriate.
For lead generation, trace the route from ad interaction to qualified or converted lead. If offline imports are intermittent, the campaign may optimize against whichever early events arrive consistently. Record import coverage and latency. A technically valid upload with sparse business coverage can still be a weak training signal.
Layer 3: audit delivery with the reporting now available
Start with channel performance, not assumptions
Performance Max is no longer accurately described as a campaign with no channel-level visibility. Google's channel performance report shows how a campaign serves across Search, Display, YouTube, Discover, Maps, Gmail, Search partners, product-data formats, and video formats. It supports date ranges after June 6, 2025 and includes channel diagnostics and conversion-goal segments (Google Ads Help: Channel performance report).
Use the report to ask:
Is delivery concentrated in the channel expected for the campaign's job?
Did the channel mix change when performance changed?
Are conversion categories materially different by channel or format?
Is a channel under-serving because required assets are missing?
Does product-data delivery account for a large share of value?
Do not automatically pause or celebrate a channel because its attributed CPA differs. Cross-channel attribution means a channel can assist a path without looking efficient in isolation, and aggregate asset or channel metrics may not support causal conclusions. Treat the report as a diagnostic map that tells you where to investigate.
Audit search demand and brand capture
Review search-term insights, search categories, brand terms, competitor terms, and irrelevant themes. Then compare the observed demand with the campaign's assigned role.
Current Performance Max search controls include Final URL expansion, search themes, campaign-level negative keywords, and brand exclusions. Google notes that Performance Max negative keywords apply to Search and Shopping inventory, not universally across every channel (Google Ads Help: Search targeting and controls; Google Ads Help: Brand suitability features).
Choose the control according to the failure:
Brand exclusion: use when branded traffic should be separated from the campaign's role. Document whether the excluded brand includes misspellings, related brands, and relevant URLs.
Negative keyword: use for specific queries that should not trigger Search or Shopping inventory. Do not describe it as a cross-channel exclusion.
Search theme: use as a signal that helps the system understand relevant demand; it is not equivalent to an exact keyword target.
Final URL controls: use URL exclusions, page feeds, or relevant URL rules when the system reaches pages that are unsuitable for acquisition.
Brand capture requires special care. A strong campaign-level ROAS may partly reflect people already seeking the company. The right action depends on the account: some businesses deliberately include branded demand; others separate it to understand non-brand acquisition. The companion guide on Google Ads brand-bidding incrementality provides a test design rather than a universal “always bid” or “never bid” rule.
Audit product data for ecommerce
For retail campaigns, the Merchant Center feed is part of the media system. Review item eligibility, disapprovals, price and availability consistency, product identifiers, titles, images, product types, custom labels, and landing-page quality. Segment results by product economics rather than relying only on campaign totals.
Ask four questions:
Are high-spend products actually in stock and commercially attractive?
Do product titles and attributes describe the item accurately enough to match relevant demand?
Are low-margin, high-return, or strategically excluded products receiving budget?
Does campaign segmentation reflect meaningful budget or economic differences, or merely create smaller data pools?
Avoid a reflex to split campaigns by every category. Segmentation is justified when the business needs different budgets, targets, countries, customer goals, or economics. An elaborate structure without a distinct decision boundary can fragment learning while adding no useful control.
Audit asset groups and creative evidence
Asset groups should have a coherent theme, audience, product set, landing-page promise, and creative proposition. Google defines them as collections of assets organized around a theme or target audience (Google Ads Help: Build an asset group). Review whether each group tells one story instead of functioning as an arbitrary replica of a Search ad group.
Current asset reporting can include conversion metrics in asset-group, asset, and asset-association reports. However, Google cautions that when one ad containing multiple assets converts, each asset can receive credit; asset-level values therefore should not be summed to reproduce the asset-group total (Google Ads Help: Conversion reporting by asset).
Use asset data to form hypotheses, not to declare a single image causal. Evaluate:
coverage across the formats the campaign needs;
alignment between creative promise and landing page;
product and audience specificity;
repeated concepts disguised as “variety”;
whether generated or customized assets remain factually and legally acceptable;
whether low-volume assets simply lacked delivery.
A useful creative refresh changes a proposition, proof type, use case, or objection—not just a background color. Preserve the original set long enough to compare the change under a defined window.
Audit inventory and brand suitability
Review placement reporting and the account's content-suitability controls. Account-level placement exclusions can apply to specified Display, Search partner, and Video placements, while campaign and account controls have different scopes. Google also warns that exclusions can reduce reach and that no content-exclusion system guarantees every related placement will be blocked (Google Ads Help: Brand suitability features).
The audit should separate three concerns:
brand safety: content the organization cannot appear beside;
brand suitability: content that is permitted but misaligned;
performance: inventory that appears inefficient in attribution reports.
Do not use a placement exclusion as a causal performance conclusion when the report cannot establish incrementality. Use it when there is a documented suitability requirement or a sufficiently supported operational reason.
Layer 4: test incrementality and channel fit
Platform attribution answers which interactions received credit under the platform's model. Incrementality asks what happened because the campaign ran. These are different questions.
Google provides Performance Max experiments for eligible cases, including testing uplift from adding Performance Max and testing shifts from certain existing campaign types. Google recommends a business-linked hypothesis before creating an experiment (Google Ads Help: About Performance Max experiments; Google Ads Help: Test with confidence). Availability and eligibility vary by account, so confirm what the interface supports rather than assuming a specific test is present.
Where a native experiment is not suitable, consider a carefully designed geo, audience, product, or time-based test only when contamination, seasonality, and sample size can be managed. The paid media incrementality testing guide explains how to choose among holdouts, platform experiments, matched markets, and broader measurement methods.
Define the experiment record before launch:
hypothesis;
treatment and control;
primary business metric;
guardrail metrics;
expected conversion delay;
minimum observation rule;
contamination risks;
decision thresholds;
conditions that invalidate the test.
Do not invent a universal number of conversions or weeks. The needed evidence depends on baseline volume, variance, conversion lag, effect size, seasonality, and the cost of a wrong decision.
A practical audit workflow
Step 1: freeze and document the baseline
Record the date range, budget, bid strategy, targets, goals, customer setting, brand controls, URL behavior, asset groups, product filters, and material recent changes. Export the relevant reports. Avoid making simultaneous edits while diagnosing a trend.
Step 2: identify the first broken layer
Work in order:
business outcome;
conversion measurement and values;
customer classification;
feed and landing-page integrity;
delivery mix and search demand;
assets and suitability;
incrementality.
If an upstream layer is broken, repair it before interpreting downstream optimization. A campaign cannot produce trustworthy value-based decisions from untrustworthy values.
Step 3: make one decision-grade change
Examples include removing a duplicate primary conversion, importing qualified-lead outcomes, excluding the house brand, correcting a feed subset, constraining URLs, or replacing a materially different creative proposition. State what observation would support or reject the change.
Step 4: wait for the outcome, not just the interface
Allow for conversion lag and the campaign's response to a significant change. Google's guidance notes that learning after significant changes can require time, but the appropriate observation window still depends on the account's volume and business delay (Google Ads Help: Channel performance report). Do not restart the campaign repeatedly because early results are noisy.
Step 5: issue the decision memo
The memo should fit on one page:
campaign role;
target customer and economic outcome;
measurement confidence;
material findings by layer;
decision: keep, fix, constrain, or replace;
change to make;
test and observation window;
owner;
next review date;
known limitations.
This converts the audit from a list of interface settings into an accountable operating decision.
When Performance Max may be the wrong tool
Performance Max may be a poor primary campaign type when the business requires exact query-level budget control, a tightly prescribed placement plan, a buying motion that cannot send meaningful outcomes back to the platform, or a regulated approval process incompatible with dynamic asset combinations. Google's channel-reporting guidance explicitly suggests standard Search alongside Performance Max when Search traffic with more manual control is the primary goal (Google Ads Help: Channel performance report).
It may also be premature when conversion volume is sparse and the available event is far from value. That does not imply automation is universally bad. It means the current signal and control requirements do not support the job assigned to the campaign.
Conversely, do not replace Performance Max merely because it is automated. If its goals are credible, delivery fits the role, customer economics reconcile, and a controlled comparison supports its contribution, removing it can sacrifice useful reach without improving decision quality.
Questions growth teams should ask during the audit
Is Performance Max taking credit for branded searches?
It can serve on branded queries unless relevant controls are applied. Review search-term insights and brand traffic, define whether brand belongs in the campaign's role, and use brand exclusions or an experiment where appropriate. Do not assume all branded conversions are non-incremental.
Can Performance Max negative keywords block all unwanted inventory?
No. Google states that Performance Max negative keywords apply to Search and Shopping inventory. Content suitability and placement controls address different inventory and operate at campaign, account, or manager-account scope depending on the control.
Should every low-performing asset be removed?
No. Some eligible assets receive little delivery, and asset-level conversion credit is not additive. Replace assets when the evidence and creative strategy support a new hypothesis, not because one column appears low in isolation.
Should ecommerce teams separate new and returning customers?
They should at least measure and discuss the mix. Whether to use a lifecycle goal or separate economics depends on the reliability of customer identification, the campaign's role, and the business's retention model.
What is the strongest reason to pause Performance Max?
A pause is strongest when measurement is credible, the campaign's intended role is clear, a fair observation window has passed, and a controlled test or reconciled business result shows that the budget does not produce acceptable incremental economics. Discomfort with automation alone is not a decision rule.
The next action
If you lead growth for an ecommerce, app, SaaS, or lead-generation business and Performance Max reports look stronger than the economics—or the team cannot explain where the campaign creates value—a focused audit may be appropriate. Sharply Labs' growth marketing work can examine conversion goals, value flow, brand and search controls, channel delivery, feed or asset-group logic, and the campaign's role alongside the rest of paid acquisition.
The output should be a prioritized keep, fix, constrain, or replace decision plan with measurement gaps and testable next steps. It is not a promise of a particular ROAS, CAC reduction, or incremental lift; those outcomes require account-specific evidence.