PPC Audit: What To Check And Where To Look

An account audit is a sequence of checks. Each one has a place to look, a control behind it, and a documented behavior you can hold the account against. That’s a different thing from an opinion about whether your campaigns are structured well.

When someone tells you your account is too granular, there’s nothing behind the sentence you can verify. When someone tells you your top Search campaign is running at 40% impression share with a quarter of the loss attributed to budget, you can open the same report and read the same numbers. The second kind of finding survives a meeting. The first one doesn’t.

Every check below points at Google’s own documentation for the control it describes, with the link in the sentence that uses it. That means nearly every source here is a vendor describing its own product, so where Google’s documentation is also a marketing position, we say so in the sentence. These pages were read in September 2026 and Google revises them, so re-read the one your decision turns on.

We run paid media in-house at BLKDG, and this is the order we work in when we open an account we didn’t build.

Google Ads is organized into three layers: the account, campaigns, and ad groups. Google’s account organization documentation (opens in new tab) describes campaigns as carrying their own budget and settings that determine where your ads appear, and ad groups as containing a set of similar ads and keywords. Every check in an audit resolves to a setting on one of those layers, or to a report built across them.

That’s the standard a finding has to meet. A named report, a named setting, or a behavior Google documents. Anything else is a preference.

Google doesn’t publish a doctrine on consolidation versus granularity. The account organization article describes the three layers and stops there, so there’s no Google-documented standard that says a 40-ad-group campaign is better or worse than a 4-ad-group campaign. What Google does publish is hard limits and serving-priority rules, and you can reason about structure from those instead.

How often should an account be reviewed? The mechanics point at events rather than a calendar. Conversion counting changes and conversion window changes apply only going forward, bid strategy changes restart a learning period, and auto-apply enrollment is recorded in change history. So the useful triggers are a bid strategy change, a conversion setup change, or an agency transition, and the first thing to read in all three cases is the change history.

Google describes Quality Score as a diagnostic tool meant to give you a sense of how well your ad quality compares to other advertisers, measured 1 to 10 and available at the keyword level. The Quality Score documentation (opens in new tab) then says two things that remove it from the audit entirely.

Cite: Google Ads Help, About Quality Score for Search campaigns

The same page states that Quality Score is not a key performance indicator and should not be optimized or aggregated with the rest of your data. An account-level average Quality Score is, by definition, an aggregation. A finding that reads “raise account Quality Score from 6.2 to 7.5” is proposing work against a number the vendor says isn’t an auction input and shouldn’t be aggregated in the first place.

Two more details from the same page change how you read the column. Quality Score is based on historical impressions for exact searches of your keyword, so changing keyword match types won’t move it. And a dash in the Quality Score column means there aren’t enough searches that exactly match the keyword to determine a score, which is a volume statement rather than a quality statement.

Scope this one carefully. That article is titled for Search campaigns, so the 1-to-10 keyword-level score doesn’t carry over to Shopping or Performance Max, and an audit that reports a Quality Score finding on a PMax campaign is reporting something that doesn’t exist there.

Google defines optimization score as an estimate of how well your Google Ads account is set to perform, running from 0 to 100%. That’s the vendor’s own framing of its own product, and the word Google uses is estimate. The optimization score documentation (opens in new tab) says it’s calculated in real time from your statistics, settings, account and campaign status, the relevant impact of available recommendations, and recent recommendations history.

Then it says how the number moves: applying or dismissing these recommendations changes the overall optimization score of your account.

Dismissing a recommendation changes nothing about how your campaigns serve. No bid moves, no keyword is added, no budget shifts. The score moves anyway, because the score is partly a function of the state of your recommendations queue. A number that responds to you disagreeing with Google is measuring your agreement with Google, not the quality of your account.

Google’s API documentation (opens in new tab) describes the same metric the same way, as an estimate of account performance available at the Customer and Campaign levels. The help page adds the mechanics: each recommendation carries a score uplift from below 0.1% to 100%, reflecting the estimated impact on campaign metrics, and the score is shown only for active Search, Display, Video Action, App, Performance Max, Demand Gen, and Shopping campaigns.

Read individual recommendations on their merits. Some of them are good. The composite is a vendor metric and doesn’t belong in a board deck as a measure of account health.

Google documents auto-apply as something the advertiser turns on. The auto-apply article (opens in new tab) says that when you turn on “Automatically apply recommendations,” the recommendations will apply regularly, that only a select number of recommendations can be automatically applied, and that recommendations are applied only when they’re relevant, so some may be applied frequently and others may never be applied at all. It also states that auto-applying recommendations won’t increase your budget.

None of Google’s auto-apply pages states whether the feature is enabled by default for a new or an existing account. Every one of them uses opt-in language throughout. Google documents neither state, so neither will we.

The check that works without knowing the default is better anyway. Google’s auto-apply management article (opens in new tab) documents a confirm-opt-in procedure and states that you can confirm who enrolled your account, and when, through the account’s change history, and that you can also confirm the user ID that opted in. Open the Recommendations page, then auto-apply settings and the History tab, then Change history for the user ID.

You’ve found a defect when the enrolling user ID belongs to nobody on the current team, or when the enrollment date predates your engagement and nobody inherited the decision. Whether to stay enrolled is a judgment call. Whether you know you’re enrolled isn’t.

Google also documents a bundle called Maintain Your Ads (opens in new tab), described as a starter pack of recommendations that keeps essential Google Ads best practices in place, and says opting in lets Google apply fundamental keyword, targeting, and ad optimizations automatically. Google states this set will never increase or change your budget. It will change keywords and targeting, which is the part to price into the decision.

The structural limits are published, and they tell you where the real constraints sit. Google’s account limits page (opens in new tab) documents 10,000 campaigns per account including active and paused, 20,000 ad groups per campaign, 100 Performance Max campaigns per account, and 25 maximum campaign drafts, with Local and App campaigns limited to 100 ad groups per campaign.

The same page caps targeting items at 5 million ad group targeting items per account, covering keywords, placements, and audience lists, and 1 million campaign targeting items per account, covering location targets and campaign-level negative keywords. Doing our own arithmetic on those two numbers: 10,000 campaigns times 20,000 ad groups is a theoretical 200 million ad groups, which the 5 million targeting-item cap makes unreachable by a factor of forty. Keywords and audiences run out long before containers do, so the granularity argument almost never hits a real ceiling.

The location number has a trap in it. Google’s 10,500 location targets per campaign, split into 10,000 individual and 500 proximity, explicitly includes both targeted and excluded locations. A campaign excluding 2,000 ZIP codes has already spent nearly a fifth of its allowance on exclusions, which matters for accounts running tight geographic carve-outs.

The 100 Performance Max campaigns per account limit is the one that binds first in practice, because a PMax-per-category structure scales linearly with catalog breadth. Where to look: the campaigns table filtered by type, counted against 100.

Google withholds part of the report and documents that it does. The search terms report page (opens in new tab) states that some search terms that don’t have enough query activity are omitted from the search terms report in order to keep with Google’s standards on data privacy, and describes the report as a list of search terms that a significant number of people have used. Search terms insights account for the omitted queries by grouping them into relevant subthemes where applicable, or aggregating them as other queries, without exposing the queries themselves.

Google publishes no number for “enough query activity” or “a significant number of people.” Any specific figure attached to the size of that omission is invented.

The consequence for the audit is a bounded claim rather than a total one. You can’t audit search terms to zero, because the long tail is structurally invisible, so negative keyword coverage is always measured against the visible portion. An auditor who promises to eliminate all irrelevant queries is promising something the report can’t support.

The 56-day figure on that page belongs to a narrower noun than it’s usually attached to. Google says labels for the themes and subthemes are generated from the last 56 days of data, so they may contain terms that are no longer serving or are excluded. That’s the label-generation window for search terms insights, not a retention limit on the search terms report itself.

The match type column needs the same care. Google notes that a search term match type listed in your report might not be the same as the match type you’ve selected for the keyword that triggered the ads, because keywords with broader match types can still match search terms in narrower ways. And terms matched to Dynamic Search Ads or Shopping targets return no keyword in the keywords field while returning Exact in the match type field.

So a screen full of Exact rows with a blank keyword column isn’t a tight exact-match build. It’s Dynamic Search Ads or Shopping, and the exclusion controls for those are different from the ones you’d reach for.

Negative keywords don’t behave like positive keywords, and the difference is where leakage comes from. Google’s negative keywords documentation (opens in new tab) explains that negatives don’t match close variants or other expansions: exclude the negative broad match keyword “flowers” and your ads won’t be eligible on a search for “red flowers”, but they can still serve on “red flower”. You have to add synonyms and singular or plural forms yourself. Casing and misspellings are handled automatically, so you don’t need to add those separately.

That’s a maintenance property, not a one-time build. A negative list assembled once, without plurals, leaks from the day it ships. Negative broad match is the default type, and account-level negative keyword lists apply automatically to eligible search and shopping inventory in relevant campaign types.

Now the limit that gets misquoted most often. Google’s account limits page (opens in new tab) documents 10,000 negative keywords per campaign, 5,000 keywords per negative keyword list, and a maximum of 20 lists for a manager account with another 20 available to each child account. Separately, it caps application at 1,000 negative keywords for Display Network and Video campaigns, and the negative keywords article states the same scoping independently, describing a maximum of 1,000 negative keywords considered at the account level for Display and Video ads.

The 1,000 figure is Display and Video. Search campaigns get 10,000 per campaign, ten times as many, and an audit that reports “Google caps you at 1,000 negatives” against a Search campaign has attached a real number to the wrong campaign type.

Our own arithmetic on list capacity: 20 lists at 5,000 keywords each is 100,000 negatives available to a manager account, against a 10,000 application ceiling per Search campaign. Storage isn’t the binding constraint. If a Search campaign is near 10,000 applied negatives, the problem is usually match-type strategy upstream, because a broad-match build generates exclusions faster than any list can absorb them.

Percentages of wasted ad spend circulate widely, usually as a single headline figure attached to the average account. We went looking for a primary source for the most-cited version of that claim and didn’t find one. Two candidate origin URLs return 404, and the publisher’s own index carries no such figure. We’re not going to quote a number we can’t trace back to a document.

What actually counts as waste? Three things, each of which has a report behind it and a dollar figure you can compute from your own account.

The first is spend on search terms you’d exclude if you saw them, bounded by the visible portion of the search terms report. The second is conversions counted more than once, which understates cost per conversion and pulls budget toward whichever campaign double-counts hardest. The third is auctions you were eligible for and skipped because the campaign hit its budget, which Google reports as lost impression share attributed to budget.

Each of those produces a defensible number sourced to a named report. That’s a stronger thing to put in front of a CFO than an industry average, because it’s yours and because the report is one click away if anyone challenges it.

Every ROAS figure, every cost per acquisition, and every Smart Bidding decision is computed from conversion data. A defect here doesn’t produce one wrong number, it produces a consistent bias across the entire account, and it biases the bid strategy in the same direction it biases the report. That’s why the measurement layer gets audited before anything that depends on it.

Google’s documentation on “All conversions” (opens in new tab) states that the column includes the data in your “Conversions” column, conversion actions you’ve chosen not to include in that column, store visits, certain phone calls, and more. View-through conversions are not included in the “Conversions” column, appearing only in the “View-through conversions” and “All conversions” columns.

Calls split across the two columns by device. Conversions generated by calls from mobile search ads are counted as standard conversions and included in the Conversions column, while calls placed after viewing ads on tablets and computers are included in All conversions.

A report built on All conversions reads higher than the same report built on Conversions, and the gap isn’t an error. It’s the column definition. Where to look: the column header in whatever monthly report the account has been sending. When the header doesn’t say which column it is, the number can’t be reconciled against anything.

Google’s primary and secondary conversion actions page (opens in new tab) defines primary actions as the ones reported in the “Conversions” column and used for bidding, as long as the standard goal they’re part of is used for bidding. Secondary actions are for observation only, reported in “All conversions” but not used for bidding even when the goal they belong to is used for bidding, with one exception for secondary actions inside a custom goal.

Then the default that catches accounts out: these conversions are secondary by default when created from Google Analytics and can only be updated to primary in Google Ads.

An account that imported its GA4 key events and expected them to drive Smart Bidding is misconfigured by default rather than by mistake. Nobody made an error. The default did it, and it’s invisible unless you open the conversion actions table and read the goal column.

Google calls correct primary and secondary configuration critical and says misconfiguring these settings can prevent Smart Bidding from optimizing effectively. The same page also notes that primary conversion actions not used for optimization may still be used to enhance predictions.

If the import itself is in question, Google’s conversion measurement page (opens in new tab) lists the prerequisites: Admin access in Google Ads and Edit access in the linked Google Analytics property, only events marked as key events in Analytics are eligible for import, and after linking accounts or marking an event as a key event it can take 24 to 48 hours to propagate before the import options appear in Google Ads. A team that linked accounts yesterday and found nothing to import hasn’t hit a bug.

Google’s conversion counting documentation (opens in new tab) gives you two settings per conversion action. Every conversion counts every conversion per tracked action that happens after an ad interaction, which Google recommends for sales because every sale likely adds value. One conversion counts only one conversion per ad click, which Google recommends when you care whether a certain kind of lead was generated rather than how many.

Google names the failure mode itself. Its conversion tracking guidance (opens in new tab) says to select “Every” for sales to track each purchase and “One” for leads to prevent duplicate counts from page refreshes, and to use “Transaction IDs” to deduplicate sales and ensure only bottom-of-funnel actions are set as “Primary” to avoid inflating counts. That single sentence contains the mechanism, the remedy, and the configuration rule.

The magnitude is computable from a metric Google already publishes. Repeat rate is defined as the number of conversions you’d have recorded using the every-conversion setting divided by the number you’d record using the one-conversion setting. Running our own arithmetic on that definition, with an illustrative repeat rate of 1.5: a 900-conversion month describes 600 interactions that converted at least once, and a reported $40 cost per conversion is $60 per unique lead.

That gap is the difference between a channel that clears a $50 target and one that misses it, on identical spend and identical traffic. The bid strategy sees the same inflated number you do.

One more property closes the loop. Google states that any changes to conversion count settings will only apply to future conversion reporting, so you can’t repair history. You can only date the break and annotate the reporting period on either side of it.

Tracking defects that arrive silently are their own category of problem. Conversion tracking that stops reporting without throwing an error is the same class.

In the accounts we open, a lead-form action set to “Every” with no transaction ID is the finding we hit most often.

A conversion window is the period after an ad interaction during which a conversion is recorded in Google Ads, and Google’s conversion windows page (opens in new tab) describes a default 30-day window. Set it to 7 days and any conversion that happens more than 7 days after the interaction won’t be recorded, which means it won’t appear in your reports at all.

Changes move forward only. Google states that if you’re using a 30-day window and change it to 10 days, the 10-day window applies only to conversions recorded from that day forward. Its worked example goes further: change the window again to 20 days on March 16, and a March 13 conversion from the original interaction that wasn’t counted under the previous window won’t be retroactively counted.

So a year-over-year comparison that spans a window change compares two different measurement rules, and part of the delta is the rule. An audit that reports a 22% conversion decline without checking the window change log may be reporting a settings change as a performance change.

Two more measurement-layer checks belong here. Enhanced conversions (opens in new tab) supplements existing conversion data by sending hashed first-party data from your website tags or imported offline events, using SHA256, in hex. Google documents two variants: enhanced conversions for web, which sends user-provided data from your site at the moment of conversion and matches it to signed-in Google accounts, and enhanced conversions for leads, which pairs hashed lead form data with imported offline lead conversions.

Tag health has named states. Google’s troubleshooting page (opens in new tab) points you to troubleshooting when a tag status shows as Inactive, Unverified, or Needs Attention, and tells Google Tag Manager users to confirm the Conversion Linker tag is set up correctly. A primary conversion action reading Unverified while a bid strategy optimizes toward it is a defect with a dollar value attached. If the tag layer itself is unfamiliar territory, our introduction to Google Tag Manager covers the container setup these statuses depend on.

Four models are gone. Google’s attribution models page (opens in new tab) states that the first click, linear, time decay, and position-based attribution models are no longer supported, that conversion actions using the deprecated models have been upgraded to data-driven attribution, and that you can also switch to the last click model, which is still supported. Two models remain: last click, which gives all credit to the last-clicked ad and corresponding keyword, and data-driven, which distributes credit based on past data for that conversion action and is the default for most conversion actions.

That makes account documentation auditable. An SOP, a reporting deck, or an onboarding doc still specifying position-based or time-decay attribution is describing something that doesn’t exist in the product. It dates the account’s governance precisely, and it usually predicts what else hasn’t been revisited.

The data-driven threshold is the one most likely to be misquoted, because the wrong version used to be right. Google’s data-driven attribution page (opens in new tab) opens by saying all conversion actions are eligible for data-driven attribution regardless of conversion or interaction volume. It then says the performance of the model improves with more data and recommends having at least 200 conversions and 2,000 ad interactions in supported networks within a 30-day period.

Those are two different claims. Eligibility is universal; 200 conversions and 2,000 interactions is a quality recommendation.

Older Google documentation did carry a hard volume gate, and it used these exact numbers: falling below 200 conversions or 2,000 ad interactions in 30 days used to switch a conversion action off the model. That’s why “DDA requires 200 conversions” still sounds correct to people who learned it when it was. An audit that recommends switching a low-volume account off DDA for eligibility reasons is recommending work against a rule the current page contradicts in its first sentence.

Attribution isn’t a reporting-only setting. The attribution models page states the setting also affects any bid strategy that uses the data in the “Conversions” column, naming Target CPA, Enhanced CPC, and Target ROAS. Data-driven attribution looks at website, store visit, and Google Analytics conversions from Search including Shopping, YouTube, Display, and Demand Gen. Where a channel sits outside that list, the model isn’t seeing it, which is the boundary worth understanding before you extend the conclusion across multi-touch attribution and the tools that stitch it together.

Google publishes the formula: impression share equals impressions divided by total eligible impressions. It also publishes the caveat that eligible impressions are estimated using many factors including targeting settings, approval statuses, and quality, and that impression share (opens in new tab) is based on an estimate of when your ad was competitive in the auction.

The denominator is bounded by plausibility rather than by every query that could have matched. Google’s example is that impression share could include auctions where your ad could show at twice its current bid, while excluding auctions estimated to need a 1,000% bid increase to appear. Google also notes that small fluctuations don’t necessarily indicate action is needed, and that changes to bids, quality, or Google’s ad systems may change which auctions the system estimates you were competitive in.

Our own arithmetic on the published formula: a campaign with 12,000 impressions at 40% impression share implies roughly 30,000 eligible impressions and about 18,000 that never served. That’s a volume statement you can size a budget case against.

Here’s where the two metric families part company, and it changes how you’re allowed to combine them. Google’s impression share metric definitions (opens in new tab) define Search lost IS (budget) as the percentage of time that your ads weren’t shown on the Search Network due to insufficient budget, and Search lost IS (rank) as the percentage of time they weren’t shown due to poor Ad Rank in the auction.

Impression share is defined over impressions. Lost impression share is defined over time. Google doesn’t reconcile the two denominators anywhere on those pages, so multiplying a lost IS percentage by eligible impressions produces a number that isn’t defined in the documentation, however reasonable it looks in a slide.

Lost IS (budget) is available at the campaign level only, which Google states for both the Search and Display versions of the metric. An audit reporting lost IS (budget) per ad group is reporting a figure that doesn’t exist at that level.

The suppression note is the one that produces wrong conclusions rather than missing ones. Google states that lost IS (rank) won’t be shown on your Ad groups tab if you ran out of budget at any point during the date range being examined. A budget-constrained account shows blanks in the rank column, and a blank reads as “no rank problem” to anyone scanning quickly. It means the opposite: the account was capped, so Google suppressed the comparison.

Report Editor has its own boundary. Search lost IS (budget) is available in the Campaigns report but not the Ad group report, and as of June 2022 Google removed the metric from previously saved reports that also used Ad group, Search Keyword, or Label (Ad group). Search exact match IS is unavailable for Shopping campaigns. And on the impression share page itself, Google notes that search partners don’t distinguish between top and other positions and aren’t included in impression share data at all.

Auction insights (opens in new tab) is available for Search, Shopping, and Performance Max campaigns, and what it returns depends on which one you’re looking at. Search provides six statistics: impression share, overlap rate, outranking share, position above rate, top of page rate, and absolute top of the page rate. Shopping provides three: impression share, overlap rate, and outranking share. Performance Max provides auction insights on Search and Shopping ads served on the Search Network, with data running from November 2021, segmented by Search and Shopping ads, at the account and campaign level.

The report also has a floor. Google states that auction insights doesn’t show insights when impression share is less than 10%. A small account finding the report empty is looking at documented behavior, not a broken report, and an auditor who reports “no competitive data available” without naming that threshold has skipped the explanation.

Daily spend above the daily budget is expected behavior with a published limit. Google’s bid and budget page (opens in new tab) explains that because traffic fluctuates, Google may allow a campaign to spend more in one day than the average daily budget specifies, and calls this overdelivery. It then bounds it: campaign spend will never exceed 2 times your average daily budget on a given day, and in a given billing period you’re never charged more than 30.4 multiplied by your average daily budget. Google’s own example is that a $10 per day budget tops out at $304.

A second Google page (opens in new tab) states the same 2x daily ceiling and adds that an overdelivery credit is applied to your account if your ad is shown so much that you exceed your monthly charging limit.

Running the multiplier at real budget levels: a $500 daily budget carries a documented monthly ceiling of $15,200 and a possible single day of $1,000. A finance team that approved $15,000 a month and sees $15,200 on the invoice is looking at the documented behavior of the product. That conversation goes better before the invoice than after it.

The same arithmetic gives you the setting. If $15,000 is the number you can defend, the daily budget is $15,000 divided by 30.4, which is $493.42, not $500. Google also recommends reviewing budget settings so you’re comfortable spending up to 2 times your average daily budget before applying a Target ROAS strategy.

Google’s Target ROAS documentation (opens in new tab) publishes seven conversion requirements covering eight campaign types. The commonly quoted “15 conversions” figure is correct for three of them: Search, Shopping, and Display.

Google adds that conversions must have a value greater than 0 to count as eligible, and that you need to set values for the conversions you’re tracking before you can apply a Target ROAS strategy at all. Google also notes that new Display campaigns no longer require a history of conversions to use Target ROAS.

The Demand Gen row rewards a second reading. Its first branch is compound: 50 conversions in 35 days, of which at least 10 must have landed in the past 7. An account can clear the 50 comfortably on a strong month six weeks ago and still fail the recency leg, which makes that eligibility intermittent rather than a gate you pass once. Any audit that treats bid strategy eligibility as a one-time check will misread a seasonal account.

The strategies themselves are named on Google’s Smart Bidding page (opens in new tab): Target CPA, Target ROAS, Maximize conversions, and Maximize conversion value, all using auction-time bidding. Google’s note states that starting in June 2026 it updated how bidding strategies are labeled, with “Maximize conversions with a Target CPA” becoming “Target CPA” and “Maximize conversion value with a Target ROAS” becoming “Target ROAS,” while the underlying bidding behavior stays exactly the same. That change is already in effect, so an internal document still using the longer labels predates it, which dates the document.

Google describes the learning period in two different units on two different pages, and they shouldn’t be merged into one claim. The learning period article (opens in new tab) says it can take up to around 50 conversion events or 3 conversion cycles for the bid strategy to calibrate to the new objective, and that the duration depends mainly on conversion volume, conversion cycle length, and the bid strategy. The same page notes it isn’t applicable to Manual CPC, and that Google’s algorithms keep learning after the Learning status clears.

A separate note on the conversion measurement page (opens in new tab) says Smart Bidding strategies require a standard 7 to 14 day learning phase and that frequent manual changes to budgets, targets, or conversion goals will reset the learning window and delay optimization.

The reset mechanism is the audit-relevant half. Google documents three triggers for a Learning status: a setting change on the bid strategy, a composition change where campaigns, ad groups, or keywords are added or removed from the strategy, and in some cases an ad group target change on Shopping campaigns. An account with three target changes inside ten days has been re-entering learning rather than progressing through it, and the fix is a change freeze rather than a new target.

Shared budgets change what any of this means at the campaign level. Google’s shared budget documentation (opens in new tab) says you can assign an average daily budget to each individual campaign or share one across any number of campaigns, and notes that some accounts can now link portfolio bid strategies and shared budgets together. Since lost IS (budget) is reported at the campaign level while a shared budget is consumed across campaigns, an account on shared budgets can show several campaigns losing impression share to a constraint none of them owns.

Google publishes a search terms report for Performance Max. Its documentation (opens in new tab) describes the report as showing the search terms that triggered your Performance Max ads, found under Insights and reports in the Campaign menu, then Search terms and landing pages for Performance Max. Historical data runs back to March 2023 and earlier data isn’t shown.

The report segments in two ways that matter for an audit. Ad format segmentation shows the ad type served for each search term, distinguishing Shopping ads from Text ads, and the report pairs terms with landing pages. Store visits and store sales conversions aren’t available in it.

So the search terms check on a PMax campaign is the same check you run on Search: read the terms, find the ones you’d never have bid on, exclude them. The controls are where it diverges.

Google documents two different exclusion mechanisms for two inventory types inside a single campaign. For Search and Shopping inventory, you apply negative keywords at the account or campaign level. For Display and Video inventory, you use excluded content keywords in the Content Suitability center to keep ads from showing alongside irrelevant content.

A negative keyword list doesn’t cover Display and Video inventory in a PMax campaign. An audit that checks only the negative keyword list has checked one of the two controls and will report the campaign as protected when half of it isn’t.

Serving priority against your Search campaigns is documented too. Google’s Performance Max page (opens in new tab) states that Search campaigns containing an exact match keyword are prioritized to serve over Performance Max when a query hits an exact match keyword, and that search themes carry the same priority as phrase and broad match keywords. Search themes (opens in new tab) are optional and additive to what PMax predicts from your assets, feeds, and landing pages, capped at 50 per asset group rather than per campaign or per account, and exclusions and negative keywords still apply to them.

That prioritization rule is the structural answer to PMax cannibalization questions. Exact match keywords in a Search campaign win the query. Search themes don’t, because they sit at phrase and broad priority.

Two PMax reports come with usage constraints. Google states plainly that placement reports (opens in new tab) shouldn’t be used to evaluate performance, because the information in them doesn’t include performance from all channels, and that they should be used as a brand safety tool. An audit that ranks placements by cost out of that report and recommends exclusions on performance grounds is using it against the vendor’s documented guidance.

For actual delivery analysis, Google points at the channel performance report (opens in new tab), which shows how the campaign delivers across Google’s channels and inventory and surfaces diagnostics for problem areas. Asset reporting (opens in new tab) lists each asset in the campaign and compares performance across them, and the Assets page carries a Last updated column showing creation and modification dates, which is how you date creative that hasn’t been refreshed.

One structural fact matters for anyone pulling data out programmatically. Google’s API documentation (opens in new tab) states that Performance Max campaigns don’t have standard AdGroup and AdGroupAd entities, organizing assets and targeting through AssetGroup resources instead, so querying ad_group or ad_group_ad won’t return any data for Performance Max campaigns. Those queries return no rows rather than an error, so a dashboard built on ad-group resources omits PMax entirely without reporting a gap.

The field-level reference draws the line more finely still. Google’s SearchTermView reference (opens in new tab) states that the view doesn’t include Performance Max data and points you to CampaignSearchTermView instead, and the CampaignSearchTermView reference (opens in new tab) states that if keyword-related segments are used, Performance Max data will be excluded from the results. So the PMax rows exist, and adding a keyword segment to the query drops them without warning.

On accounts where PMax carries the largest share of spend, that’s the campaign type missing from the report the client trusts most. Sorting this out is a standing part of Shopping and Performance Max management.

On a Shopping or PMax account, feed problems present as bidding problems. Google’s product data specification (opens in new tab) states that incorrect, inaccurate, or missing product information can cause disapprovals, limited eligibility, incorrect displays, or other issues in Merchant Center, and names the common ones: incorrect google_product_category or gtin values, missing or incorrect variant attributes such as item_group_id, color, or size, low-quality images, and conflicting data between your feed and website.

Three instructions on that page are checkable against the site in a browser. Accurately describe the product and match the title from your landing page. Accurately submit availability and match it to the landing page, checkout pages, and structured data. And if an item is out of stock, the price must still be clearly visible on the landing page.

Title is plain text with a maximum of 150 characters.

Availability is required for all products, and Google’s availability documentation (opens in new tab) states the values in your data source should always match the information on your website. Four values are documented:

Google distinguishes two of them explicitly. Preorder is only for new products that haven’t been released yet, while existing products that are out of stock and will return, where you’re accepting orders, should use backorder. Accounts running seasonal or made-to-order inventory frequently have those reversed.

The delete-versus-pause instruction carries a real cost. Google says that to temporarily stop accepting orders or hide a product, don’t delete the item, because adding an offer back after deletion takes a significant amount of time before it can show again, and directs you to the pause attribute instead. A merchandising team that deletes end-of-season products and relists them next year is paying re-serving time it didn’t need to spend.

At the account level, Merchant Center guidelines (opens in new tab) state that products promoted in Shopping need to be available for purchase through your online store, that you’re not allowed to use Shopping to promote affiliate or pay-per-click links except as a Comparison Shopping Service in a CSS program country, that you must submit different product data for each language with landing pages in the same language, and that you have to state your return policy explicitly even if you don’t offer returns or refunds.

Feed accuracy has stopped being only a Shopping concern, since the same product data now feeds surfaces beyond the ad auction. We covered that shift in how product feeds drive visibility in AI shopping surfaces. The audit check is unchanged either way: pull ten products at random, open the PDP next to the feed row, and compare title, price, and availability field by field.

Sort the findings in the order the dependencies run. Measurement first, then waste, then structure.

Measurement goes first because every other number is computed from it. Fix a double-counting conversion action and your cost per conversion, your ROAS, and your campaign ranking all change on the same day, which means any waste figure you calculated before the fix has to be recalculated after it. Waste comes second because it carries a dollar figure and a named report. Structure comes last because it’s slowest to move and because Google publishes no standard to argue it against.

Sequence the fixes rather than shipping them together. Changing conversion settings and bid strategy targets in the same week puts the campaign back into learning while the data feeding it is also changing, and you’ll lose the ability to attribute the result to either change. Fix the measurement layer, let it stabilize, then touch bidding.

That’s the audit. Not a score, not a grade out of 100, not a list of recommendations to apply. A sequence of checks, each with a place to look and a documented standard, producing findings you can still defend three months later when someone asks where the money went.

If you want that run on your account, our free Growth Audit is a no-obligation look at why you aren’t getting found or converting, with a roadmap for fixing it. No sales call. No quote request. If the wider spend picture is the question, start with paid media management instead, and if the post-click side is where the numbers break down, our ecommerce conversion rate optimization program covers the other half of the equation.

You built something worth finding. Let’s make sure the account isn’t the reason nobody finds it.