How to audit your own Google Ads account

Section 1: Where is the search-term leak?

Open the search terms report for the last 90 days, sort by cost, and read every query above 1% of spend. You are hunting four species of waste: job seekers, competitor-support queries, informational questions with no purchase intent, and geography or product mismatches. Tag each wasted query with its 90-day cost.

The total is usually uncomfortable. One HR-software account we took over had 31% of spend on terms nobody had reviewed in six months; the full mechanics are in that teardown. Your number goes into the fix plan as an annualized euro figure, because "we have some irrelevant queries" never wins budget for a cleanup and "€14,300 a year" does.

While in this report, check the invalid-traffic angle. Google filters some invalid clicks automatically, but industry measurement still put 8.51% of paid traffic as invalid in 2025, roughly one click in twelve, worth an estimated $63 billion globally. Spikes of short, repeated, zero-duration clicks from one region are worth a formal invalid-click credit request.

Section 2: Has your match-type structure drifted?

List every keyword with its match type and 90-day spend split. Google loosened matching behavior over several years, so broad match today reaches far beyond its seed keyword. Broad match is a tool, but an account where 70% of spend flows through broad keywords with a stale negative list is a drift machine.

Two structural checks matter most. First, does each ad group hold one tight intent, or has it accreted 40 keywords with three different meanings? Second, are brand and non-brand separated? Blended brand campaigns flatter every metric they touch; the industrial teardown shows a €31,000-a-month account whose real non-brand lead cost was invisible until the split.

Section 3: Is Performance Max earning its budget?

PMax reports at campaign level, which hides its favorite trick: re-buying your brand traffic. Check three things. Are brand exclusions applied? Does search still win the exact-match queries you care about, or has PMax cannibalized them? And what share of PMax conversions comes from remarketing audiences that other campaigns already fund?

Since Google Marketing Live in May 2025, Google has pushed PMax, Demand Gen, and AI Max as a package. Adopt on your evidence, never on the pitch: our Performance Max guide for B2B covers when each one earns a test and when to decline.

Section 4: Are bids set from data or from wishes?

Pull each campaign's target CPA or ROAS and compare it against the trailing 90-day actual. A tCPA of €40 on a campaign that has never delivered under €95 does no bidding magic; it starves delivery. Targets belong within roughly 20% of demonstrated performance, tightened stepwise as results improve.

Then sanity-check against your vertical. The 2026 WordStream benchmarks put median cost per lead at $66.69 across industries and $93.69 for business services, on a median CPC of $5.42. A B2B account demanding €25 leads in a $93.69 vertical is planning around a wish. The same report shows conversion rates improved in 87% of industries year over year, so a stale target from 2024 is probably also too loose.

Section 5: Do your ads still have a challenger?

Count responsive search ads per ad group and check when each last changed. One RSA untouched for six months is the norm in audited accounts, and it means no experiment is running anywhere. Every ad group needs a control and one live variation, changed one variable at a time, with the loser archived and logged.

Check asset coverage too: sitelinks, callouts, structured snippets. Ad Rank rewards useful assets, and B2B accounts leave them generic surprisingly often. Four sitelinks pointing at pricing, case material, integrations, and a demo page outperform four pointing at generic feature pages, and cost nothing to swap.

Section 6: Can you trust your conversion numbers?

Reconcile Google Ads conversions against GA4 and your CRM for the same 90 days. Gaps under 10% are normal attribution noise. Above roughly 15%, your bidding data is materially wrong and every earlier section's conclusions weaken.

EEA and UK accounts have a compliance layer on top: Google has required Consent Mode v2 since March 2024, and disabled conversion modeling for non-compliant accounts in March 2025. Verify the two v2 signals in Tag Assistant, then fix per our conversion-tracking service or the equivalent in-house work. Measurement repairs always come first in the fix sequence.

Short on time? The two-hour triage version

When a full day is impossible, three checks find the majority of the money. Read the top 50 search terms by cost for 90 days and price the obviously wasted ones; that alone usually surfaces most of the 21% median leak. Pull the exact-match impression share on your five most important keywords, because a slide there quietly explains many "the market got worse" stories. And reconcile one month of Google Ads conversions against your CRM; a gap above 15% reprioritizes everything else you were about to do.

What the triage cannot do is sequence the fixes or price the structural problems, which is where the full pass earns its day. Treat a bad triage result as the trigger for the complete audit, never as the fix list itself: accounts patched from triage findings alone tend to re-leak within a quarter because the routine underneath never changed.

Which tools help, and which just bill you?

The audit itself needs nothing paid. Google Ads, GA4, Tag Assistant, and a spreadsheet cover all six sections, which is the point: the leak is in your data, not behind a paywall. Paid tools earn a place in the ongoing routine at scale. Optmyzr, from $249 a month, automates rule-based checks like budget pacing and disapproval alerts across many accounts; Semrush, at $139.95 a month for Pro, adds competitor visibility your own account data cannot show. Neither reads your search terms with judgment, and judgment is what the weekly hour is for. Buy tooling when the routine exists and needs speed; tooling bought instead of a routine becomes a subscription-shaped excuse.

How do you turn findings into a fix plan?

Price every finding: wasted euros per year for leaks, estimated recovered conversions for measurement gaps. Divide by the hours each fix takes. Sort descending. That single spreadsheet column beats every prioritization framework we have tried, and it is the same math our €1,900 audit delivers as a sequenced plan. Most self-audits recover their day of effort within the first week of fixes.