In short: auditing an inherited account is eight blocks in two hours: access, measurement, structure, bidding, search, ads, Display/PMax and fraud, and prioritisation. Two rules make it useful: export first and change nothing while you audit, and prioritise by euros — measurement before anything else, then whatever is cheap and profitable. Deliverable: findings and a 30/60/90 plan.
You land in an account you do not know (a new client, a previous agency, a colleague who left). In two hours you have to know what is fine, what is bleeding and where to start. This script is the whole audit, with its clock, what you export and the report that comes out of it.
Before you start · What should you prepare?
- Read access as a minimum (standard is better) through your manager account; do not ask for passwords.
- Date range: last 90 days (and the previous year for seasonality).
- An audit spreadsheet with a tab per block and one for findings (finding · estimated impact in € · effort · priority).
- Quick exports: campaigns (with IS and conversions), keywords (with QS and its components), search terms over 90 days, ads (with Ad Strength), Display/PMax placements, conversions (summary), change history over 30 days.
Block 1 · Who owns the account and who has access?
Admin → Access and security: users and their levels, linked manager accounts, 2FA. Does the advertiser own the account? Is there access sitting with people who have left? Billing: who pays, spending limit.
Block 2 · Is the account measured properly?
Goals → Conversions: actions, type (primary/secondary), categories, values, counting, windows, statuses, diagnostics (consent, enhanced conversions). Segment conversions by action across campaigns: what is actually counted as a "conversion" here? Goals per campaign. A real test if you can. Typical findings: everything set as primary, L1 dominating, no values, duplicates, actions sitting at zero.
Block 3 · Is the account structured properly?
Campaigns: names, types, networks (search partners, Display expansion), locations (presence/interest), languages, brand kept separate, ad groups by intent (size, relevance), assets. Recommendations → auto-apply. Findings: a "General" campaign, brand mixed in, expansion switched on, interest targeting on a local business, auto-apply running.
Block 4 · Are bidding and budgets right?
Strategies per campaign and portfolios; targets against the real CPA/ROAS; statuses (learning, limited); lost IS by budget and by rank; budgets against actual spend; change history for targets and budgets (how often they change; who by). Findings: impossible targets, restless hands, money in flat campaigns, profitable ones held back.
Block 5 · How is Search doing?
Search terms over 90 days sorted by cost: the intent behind the top 50; existing negatives (lists, levels); a quick n-gram mine (the 20 highest-spending roots). Keywords: spend-weighted QS, the 10 highest-spending ones with QS ≤ 4, their components. Match types and duplicates. Findings: the share of spend on irrelevant terms, roots never negativised, expensive keywords with a low QS.
Block 6 · How are the ads and landing pages?
RSAs per ad group (how many, Ad Strength, Low-rated pieces with volume, date of the last change); assets (sitelinks, callouts, images); landing pages for the top 5 ad groups (speed, consistency, call to action). Findings: the eternal ad, ad groups with a single RSA, the landing page being the homepage.
Block 7 · Are there leaks in Display and PMax?
Placements report by clicks: recognisable domains and apps; anomalous CTR and CPC; existing exclusions (campaign and account level); PMax: brand exclusion, conversion goals, placements, report by channel; video: partners. Findings: the share of spend on rubbish, PMax with no brand exclusion, apps never excluded.
Block 8 · Where should you start?
Every finding with an impact estimate: affected spend × expected improvement (e.g. "35% of Display spend on farms: ~€900/month"; "irrelevant terms: 18% of Search spend: ~€1,100/month"; "measurement: not quantifiable, but it blocks everything else"). The order: measurement first (without it nothing can be judged), then whatever returns the most money for the least effort (negatives, exclusions, dangerous settings), then structure and bidding, and lastly ads and landing pages (compounding effect, slower).
The audit report: template
- Five-line summary: overall state, the three biggest problems, the estimated money at stake, the first action.
- Measurement: the current map, findings, plan.
- Structure and settings: findings and plan (with the restructure protocol if it applies).
- Bidding and budgets: a table per campaign (strategy, target, actual, lost IS, verdict).
- Search: roots to negativise, keywords to rescue/pause, QS.
- Ads and landing pages: priority ad groups.
- Display/PMax/fraud: placements to exclude, account-level exclusions.
- A 30/60/90-day plan with owners.
- Appendices: the exports.
What shouldn't you do in an audit?
- Change things while you audit (an audit is reading).
- Judge on 7 days.
- Promise results without verified measurement.
- Present 80 findings with no priority: three big ones and a plan.
💡 Ninja trick: this script is, in large part, what the Ninja Scripts Agent does every day in every account in the fleet (measurement, dangerous settings, access, anomalies) and what the QS Analyzer, SQONS and Ninja Shield calculate on their first runs (the QS X-ray, search term mining, scored placements). In accounts with the scripts installed, the two-hour audit turns into reading three spreadsheets — and becomes the week-1 "X-ray" of any new client.
What to remember
- Eight blocks in two hours: access, measurement, structure, bidding, search, ads, Display/PMax/fraud, prioritisation.
- Export first; change nothing while you audit.
- Prioritise by euros: measurement first, then the cheap and profitable.
- Deliverable: five-line summary, findings per block, a 30/60/90 plan.
End of Module 13. Module 14 is the Shopping and Merchant Center masterclass.