The Google Ads audit checklist: 15 checks with thresholds

Brand share of search spend

Segment 90 days of search spend into brand and non-brand. People searching your brand name were already looking for you, so this spend mostly buys a click you had a strong chance of getting free. Some brand defence is rational, particularly where competitors bid on your name. A quarter of your search budget is where I start asking what it's defending against.

This also inflates your blended ROAS, which is why an account can look efficient while acquiring nobody new.

Where: Campaigns, segmented by brand and non-brand keyword lists. If nobody has built that segmentation, that's finding zero.

New versus returning customer split

If most of your conversions come from existing customers, you're running a loyalty programme with an advertising budget. That may be deliberate, but it should be a decision rather than a drift, and it's usually a drift.

Where: Conversions, segmented by new versus returning, or the new customer reporting in Performance Max and Shopping.

Spend concentration by campaign

Heavy concentration isn't automatically wrong, and consolidation genuinely helps automated bidding learn. The risk is diagnostic: when one campaign is the account, you lose the ability to compare anything to anything, and a problem inside it is invisible from outside it.

Where: Campaigns, sorted by cost, 90 day window.

Search term spend with zero conversions

Pull the search terms report for 90 days, filter to terms with zero conversions, and total the spend. Some non-converting spend is the cost of discovery, which is why the threshold isn't zero. Above roughly 15% you're funding a pattern nobody is reading.

Important limitation: Google only shows search terms that cleared a privacy threshold for query volume, so the waste you can see is a floor, not a total. Treat the visible number as the minimum.

Where: Search terms report, 90 days, filtered to zero conversions, sorted by cost.

Negative keyword list coverage

Check the modification date on your shared negative lists. Broad match and automated campaign types keep widening the net, so negatives are maintenance, not setup. A list untouched for a quarter means nobody is reading the search terms report, which makes check 04 worse than it looks.

Where: Tools, Shared library, Negative keyword lists. Check the change history.

Match type distribution

Broad match combined with automated bidding can work well, and I'm not against it. But it only works with disciplined negatives and someone reviewing search terms weekly. Broad match plus a stale negative list is the most reliable way to spend money on queries unrelated to your business.

Where: Keywords, segmented by match type, with cost.

Brand exclusions on non-brand Performance Max campaigns

Without brand exclusions, Performance Max will happily serve against your own brand queries and book the resulting conversions as its own. Your Performance Max ROAS then looks excellent because it's harvesting demand your brand already created. This is the single most common reason a Performance Max campaign appears to outperform everything else in an account.

Where: Campaign settings, Brand exclusions, per campaign.

Asset group performance spread

If a single asset group dominates delivery, the campaign has stopped exploring and the others are decoration. That's sometimes the right outcome, but it means your creative testing isn't testing anything, and refreshing the losing groups won't change results.

Where: Performance Max campaign, Asset groups, by impressions.

New customer acquisition setting

Performance Max can bid differently for new customers, or report them separately. If the goal of the campaign is growth and this isn't configured, you have no way to tell whether it's acquiring anyone or re-buying your existing base. Pairs directly with check 02.

Where: Campaign settings, Customer acquisition.

Target CPA or ROAS against actual delivery

Compare each bid strategy's target to what it actually delivered over 90 days. Persistent large variance means the target is fiction: either it was set optimistically and never revisited, or the campaign cannot reach it and is throttling delivery trying. Both are fixable once named.

Where: Bid strategy report, target versus actual.

Budget-limited campaigns

A campaign capped by budget is being prevented from finishing what it started, and it will never stabilise. Two weeks is generous. If a campaign is your best performer and it's been budget-limited for a month, that's the cheapest fix in the account.

Where: Campaigns, Status column, filtered to Limited by budget.

Portfolio strategies competing with each other

Two bid strategies chasing the same inventory bid against each other, raising your own costs. It happens gradually as campaigns get added, and nothing in the interface warns you. The symptom is rising CPCs with no change in competition.

Where: Tools, Shared library, Bid strategies, cross-referenced with campaign targeting.

Primary conversion actions, and double counting

List every conversion action and check which are set as primary, which are secondary, and whether two of them describe the same event. A purchase tracked once through the Google tag and again through a GA4 import will count twice, halving your apparent CPA. This is the most common single error I find in accounts that otherwise look well managed.

Where: Goals, Conversions, Summary, grouped by action and source.

Conversion lag against your reporting window

If a fifth of your conversions arrive more than 30 days after the click and you review performance monthly, every monthly report understates recent performance and someone is making decisions on incomplete data. This is the check that most often explains why a campaign “stopped working” the month it launched.

Where: Conversion lag report, days-to-conversion distribution.

Platform-reported conversions against your own back end

Take one month and compare total conversions reported in Google Ads to actual orders or qualified leads in your CRM or back end. A gap is normal and expected, because attribution models differ and platforms claim credit generously. A gap above about 20% in either direction means you don't currently know what your advertising does.

If Google Ads reports substantially more than your back end, suspect double counting, view-through credit, or attribution overlap with your other channels. If it reports fewer, suspect tracking loss, consent-mode gaps, or conversions falling outside the window.

Where: Google Ads conversions for a closed month, beside your CRM or order data for the same period.