- Audit in dependency order. Tracking first, because every judgement after it inherits whatever that number is worth.
- Separate brand from non-brand before you look at any performance metric. Blended numbers hide almost everything worth finding.
- Time-box it. Sixty minutes forces prioritisation; four hours produces a document nobody reads.
- Score each area out of ten so the account has a baseline you can re-run in ninety days.
- Finish with exactly three fixes. An audit that ends in twenty recommendations ends in none.
Before the clock starts
Five minutes of setup makes the next hour twice as useful. Set the date range to the last 90 days with the previous period as comparison, which is long enough to smooth weekly noise and short enough to reflect the current account rather than its history.
Then set your columns once: cost, conversions, cost per conversion, conversion value, ROAS, impression share, search lost impression share due to budget, and search lost impression share due to rank. Most audits are slow because the auditor keeps rebuilding the same view.
Know the answer to one question first
What does the business actually need this account to produce, and at what cost? Without that, you are auditing against Google's definition of good rather than the client's. An account at a 3.2 ROAS is excellent or catastrophic depending entirely on gross margin.
Tracking integrity
This is the only block where a bad finding stops the audit. If conversion data is wrong, every subsequent observation is a comment on fiction.
| Check | Where | Red flag |
|---|---|---|
| Conversion actions and status | Goals, Conversions, Summary | Actions recording zero, or "no recent conversions" warnings |
| Primary versus secondary | Conversions, Goals column | Micro-conversions like page views or tel: clicks set as primary |
| Count setting | Conversion action settings | Every, on a lead generation action |
| Duplicate actions | Conversions list | Two actions with near-identical counts, usually a legacy tag plus GTM |
| Values attached | Conversions list | Blank value column while a ROAS target is running |
| Enhanced conversions | Conversion action diagnostics | Off, or on with low coverage or match rate |
| Auto-tagging | Account settings | Off, which breaks click matching and every offline import |
| Attribution and window | Conversion action settings | A 30-day window on a business with a two-month sales cycle |
The fastest sanity check is external: ask how many enquiries or orders the business recorded last month, and compare. A gap beyond roughly 15 percent means the tracking is the finding, and the remaining fifty minutes should be spent proving where it breaks rather than critiquing ad copy. The conversion tracking guide covers that diagnosis in full.
Follow the money
Sort campaigns by cost, descending. This single view answers more questions than any other screen in the interface, and it is where most of the value in an audit is found.
Work through four questions in this block:
- How much is branded? Separate brand from non-brand spend and return. Branded search converts at a rate that flatters any average it is included in. If brand sits inside Performance Max or a generic campaign, that is finding number one.
- What is the long tail doing? Campaigns spending under about 2 percent of budget rarely justify their management overhead and almost never gather enough data to optimise.
- Where is budget capped? Search lost impression share due to budget above roughly 10 percent on a profitable campaign is money left on the table.
- What changed? Open the change history. Sudden performance shifts usually have a human cause, and it is almost always in there.
Structure and match types
You are not looking for a theoretically perfect structure. You are looking for structure that prevents the account from being managed.
| Look for | Why it matters | Severity |
|---|---|---|
| Brand and non-brand mixed | Makes every performance number unreadable and misallocates budget | High |
| Broad match with thin negatives | The most reliable source of wasted spend in any account | High |
| Ad groups with 30+ keywords | Ad relevance collapses; no message matches the query | Medium |
| Single-keyword ad groups everywhere | Fragments conversion data below the volume Smart Bidding needs | Medium |
| Duplicate keywords across campaigns | Self-competition, split data, inflated cost per click | Medium |
| Paused campaigns left in place | Cosmetic, but usually a sign nobody is actively managing the account | Low |
| Location set to presence or interest | Enquiries from areas the business cannot service | Medium |
The match type distribution is worth ten seconds of arithmetic. An account that is 90 percent broad match with a negative list of eleven words is not running a keyword strategy; it is subsidising Google's discovery process.
Where the money is being wasted
Open the search terms report for the full 90 days, sort by cost descending, and add a filter for zero conversions. This is the single most valuable ten minutes of any audit, and it produces findings a client can verify with their own eyes.
You are looking for four categories: queries with clear intent mismatch, research and DIY intent, job seekers, and competitor or adjacent products you do not sell. Total the cost of everything obviously irrelevant. That figure, expressed as a monthly number, is usually the most persuasive line in the entire audit.
Check Performance Max too
Performance Max now exposes full search terms reporting, and it is frequently where the worst waste hides, precisely because most advertisers still assume they cannot see it. The negative keywords guide covers the analysis method in detail.
Then check the negative keyword lists themselves. Are shared lists applied to every campaign, or were they attached once in 2023 and never checked? Are there negatives so broad they are blocking real traffic? Both errors are common and both are quick wins.
Bidding and targets
The strategy matters less than whether it is being fed properly. Almost every bidding problem is really a data problem or a target problem.
- Conversion volume per campaign. Smart Bidding needs a steady flow to work with. A campaign generating four conversions a month on target CPA is not optimising, it is guessing with confidence.
- Target versus actual. If the target sits far below what the campaign has ever achieved, delivery will be throttled and impression share will be low. Compare the target with the trailing 90-day average.
- Recent target changes. Frequent adjustments keep a campaign permanently re-learning. Check change history for a pattern of weekly tinkering.
- Portfolio strategies. Useful for pooling data across low-volume campaigns, and worth flagging where several small campaigns each sit below threshold.
- Seasonality adjustments. Rarely used, occasionally essential, particularly for accounts with sale periods.
Also check impression share lost to rank. High values on a profitable campaign point at ad relevance, landing page experience or a target that is simply too aggressive for the auction.
Ads, assets and landing pages
Eight minutes is enough for the structural questions, which is all an audit needs.
- Coverage. Does every ad group have at least one responsive search ad, and ideally two to compare?
- Asset quantity. Are headline and description slots filled, or is Google assembling ads from four headlines?
- Pinning. Heavy pinning on every asset removes the algorithm's ability to test combinations. Some pinning is deliberate; pinning everything is usually a legacy habit.
- Extensions and assets. Sitelinks, callouts, structured snippets, and call or lead form assets where relevant. Missing assets are free real estate left unclaimed.
- Message match. Click the ads. Does the landing page say what the ad promised, and is the conversion action visible without scrolling?
- Landing page speed and mobile layout. A slow page is a bidding problem, not just a UX one.
Resist rewriting ad copy during an audit. Note the gaps and move on; copy work is an engagement, not a finding.
Score it, then pick three fixes
Score each area out of ten. The absolute number matters less than having a repeatable baseline you can re-run in ninety days to show movement.
| Area | Weight | What a 10 looks like |
|---|---|---|
| Tracking integrity | 25% | Reconciles with the CRM, values attached, offline outcomes imported |
| Wasted spend | 20% | Search terms are clean; negative lists maintained and applied |
| Structure | 15% | Brand separated, logical grouping, no self-competition |
| Bidding | 15% | Targets grounded in margin, adequate volume, stable settings |
| Budget allocation | 10% | Money concentrated where returns are proven and incremental |
| Ads and assets | 10% | Full asset coverage, sensible pinning, extensions in use |
| Landing experience | 5% | Message match, fast, obvious conversion path |
Then choose three fixes, and only three. Rank every finding by impact against effort, and take the top-left quadrant.
Red flags and what they usually mean
| What you see | What it usually means |
|---|---|
| Conversion rate above 25 percent on generic search | Micro-conversions counted as conversions, or brand traffic mixed in |
| Cost per conversion identical across every campaign | A single conversion action firing on a page everyone reaches |
| ROAS above 15 on a campaign with no product feed | Branded search inside a campaign that should not have it |
| Impression share under 20 percent with budget unspent | Targets set beyond what the auction will deliver |
| Hundreds of keywords with zero impressions | Legacy build, split data and unmanaged bloat |
| Change history empty for six months | Nobody is managing the account, whatever the invoice says |
| Conversions counted in the thousands, sales in the dozens | Count set to Every, or a tag firing on page view |
Frequently asked questions
Can you really audit an account in an hour?+
You can reliably find the three biggest problems in an hour, which is what an audit is for. A full rebuild plan takes longer, but the diagnosis genuinely does not, provided you work in dependency order rather than wandering through the interface.
What date range should I audit?+
Ninety days with the previous period as comparison. Shorter periods are distorted by weekly noise and low conversion volume; longer ones describe an account that no longer exists.
Why start with conversion tracking?+
Because every other judgement depends on it. If conversions are over-counted or misattributed, then cost per acquisition, ROAS, bidding performance and budget allocation are all measurements of a fiction, and any recommendation built on them is confident and wrong.
What is a good Google Ads account score?+
The absolute number matters far less than the direction. Most accounts arriving for audit sit somewhere in the 50s or 60s out of 100. What matters is scoring the same way ninety days later and seeing movement in the areas you chose to fix.
Should I check Google's own optimisation score?+
Glance at it, do not manage to it. Optimisation score reflects adoption of Google's recommendations, some of which increase spend without increasing profit. It is a useful prompt and a poor target.
How often should an account be audited?+
Quarterly for a well-managed account, and immediately after any significant change in tracking, site platform, campaign structure or agency. Re-running the same scorecard is what turns an audit into a trend rather than an opinion.
Where this fits
An audit is a diagnostic, not a deliverable. Its only job is to identify the smallest number of changes that will move the account, in the right order, with an honest view of what each is worth. Tracking first, waste second, structure and bidding after that.
If you would rather have a second set of eyes on it, our Google Ads consulting engagements open with exactly this process, and the free audit gives you the findings and the three fixes whether or not we work together.