The 27-Point Google Ads Structure Audit

• Final URL expansion, Search and PMax: open Campaign settings, Final URL expansion, for every active Search and Performance Max campaign. Off is correct for any campaign built on tested, converting landing pages. A bad answer is 'on' with nobody in the account able to name which pages it substituted this month.
• The one legitimate exception: large catalog accounts with no static landing page to protect and a page set current enough to trust. Final URL expansion stays on there because the substituted pages are the campaign's whole premise. Every other account type gets it switched off by default and reviewed before it goes back on.
• URL parameters survive expansion: pull a sample of ten recent substituted landing pages from the campaign's Final URL expansion report and confirm UTM parameters and click IDs still land intact. A bad answer is a substituted page with a stripped or generic URL, which breaks attribution without ever throwing an error.
• Ad group and asset-level final URLs match what the ad promises, not a placeholder. Open five ad groups at random and check the final URL field against the headline. A bad answer is a homepage URL sitting under an ad group written for one specific product.

• Exact-match brand list exists and is attached everywhere: negative keywords on non-brand Search campaigns, brand exclusions inside every PMax campaign. A bad answer is a list that lives in one campaign and not the other three.
• Brand-variant list is the second list to build: misspellings, concatenations, and the bare domain, sourced from Search Console's top-queries report filtered down to branded strings. A bad answer is a brand list containing only the exact company name, which lets a meaningful share of branded volume leak straight through the gap.
• Branded competitor terms (brand plus 'alternative', brand plus 'vs', brand plus 'reviews') are routed on purpose, either excluded from PMax and run through a dedicated Search campaign, or excluded everywhere if the account has decided not to compete on them. A bad answer is silence: nobody has made this call on purpose.
• Brand exclusion lists are dated and owned. Check the last-edited timestamp inside the shared library. A bad answer is a list untouched since account setup with no owner named to refresh it.

• Landing page report filtered to zero conversions: Campaigns, Insights and reports, Landing pages, sorted by cost with a conversions filter set to zero. A bad answer is any row spending above the account's typical cost-per-conversion with no conversions and no record of ever being reviewed.
• Search terms report swept for the trailing 30 days on every Search campaign, cross-checked against the existing negative list. A bad answer is a repeat offender: the same irrelevant term showing up month after month because nobody added it as a negative the first time it appeared.
• Placement report reviewed for PMax and Display under Insights and reports, where ads showed. A bad answer is spend landing on placements with names like games, apps, or content-farm domains the brand would never choose on purpose.
• Device segment performance checked for one device absorbing spend at a cost the account would reject if it showed up as a keyword. A bad answer is mobile or tablet carrying most of the spend with no bid adjustment and no review in the last quarter.
• Geographic report checked against the business's actual serviceable area. A bad answer is meaningful spend landing in a country or region the business does not ship to, which usually means location targeting is set to people showing interest rather than people physically present.

• The automatically created assets setting is read before it's trusted: open the account-level assets settings and review the headlines and descriptions Google generated from the site. A bad answer is the setting left on with nobody having read a single generated asset.
• Generated assets scored against brand voice and factual accuracy, not policy compliance alone. A bad answer is an approved asset stating a price, a guarantee, or a claim that is out of date or was never true in the first place.
• Auto-generated asset serving share checked in the ad strength and asset detail view. A bad answer is a generated headline outranking the assets the team wrote by hand.
• The on-or-off decision is written down with a reason attached. A bad answer is the default left untouched because nobody knew the setting existed.

• One asset group does one job: one category, one margin tier, or one objective, never a mixed bucket. Open Performance Max, Asset groups, and read the listing filter attached to each. A bad answer is a filter broad enough to pull the entire catalog into a single group.
• Asset group count matches catalog complexity, not launch-day habit. A bad answer is a single asset group running a catalog with three distinct margin tiers and no way to bid them apart.
• Every asset group clears 'Good' or better on ad strength. A bad answer is 'Poor' or 'Average' sitting unresolved past two weeks.
• Listing filters are verified against the feed, not assumed. Sample five SKUs per asset group and confirm each lands where the filter says it should. A bad answer is any of the five landing in the wrong group.
• Search themes per asset group sit in a tight band, typically three to five, checked on each group's Insights tab. A bad answer is a single group carrying ten or more themes, which dilutes the signal any one theme is meant to carry.

• custom_label_0 carries brand isolation, brand versus non-brand or flagship versus accessory line, verified in the feed export against the listing filters that reference it.
• custom_label_1 carries margin tier, high, mid, or low, based on contribution margin rather than list price, so bidding logic can eventually treat tiers differently.
• custom_label_2 carries seasonality: always-on, seasonal, or clearance, refreshed on a calendar someone follows rather than set once at launch.
• custom_label_3 carries a performance tier refreshed from real data, and custom_label_4 stays reserved for ad-hoc flags such as test groups or geo splits. Both checked for population against the full feed export. A bad answer is either label sitting blank on more than a fifth of active SKUs.
• The whole audit is dated with a next-review entry 90 days out, not filed away as a one-time project. A bad answer is a finished audit doc with no next date in it, because every check above decays as the catalog, the algorithm, and the account's own changes accumulate.

What a bad answer in each section tends to cost

Ranges from Ad-Lab structure audits across ecom accounts. Composite figures, not one specific client's results.

How to run it

1. 01

   Pull the last 90 days before you touch a setting

   Export campaign, ad group, and asset group performance, the landing page report, the search terms report, and the current feed. The audit is only as honest as the export it runs against.

2. 02

   Work the six sections in order, mark each item pass, fail, or unsure

   Unsure counts as fail until proven otherwise. Tally fails per section so you know where the account is weakest before you start fixing anything.

3. 03

   Fix the highest-cost section first, using the table above as the tiebreaker

   Ship one section's fixes at a time and hold everything else flat for a week before touching the next section, so you can tell which fix produced which result.

4. 04

   Calendar the 90-day recheck the same day you finish

   Put the next-review date in the account's own notes, not a separate doc nobody opens. The recheck only needs the items that failed plus a fresh landing page and search terms sweep.

External resources

Authoritative references we link to alongside the template. Read them before running the audit.

• Google Ads, Performance Max overviewOfficial feature reference for the asset group and campaign settings this audit walks.
• Google Ads, asset groups in Performance MaxAuthoritative spec for how listing filters and search themes route signal.
• Google Ads, brand exclusions in Performance MaxHow to apply the two-list brand exclusion structure from section 2.
• Google Ads, search-terms reportReference for the report section 3 sweeps for waste.
• Google Merchant Center, custom labelsSpec for the five custom_label fields audited in section 6.
• Google Skillshop, Performance Max certificationFree training. Worth running through once before auditing a first account.