The 12-point Google Ads audit checklist that finds wasted spend

How to run this audit without losing a day

Two rules make an audit useful rather than performative.

The first is to do it in order. Everything downstream of conversion tracking is meaningless if the tracking is wrong, so tracking comes first — and in a surprising number of accounts it ends there, because the rest of the audit turns out to be measuring a fiction.

The second is to write down what you find without fixing anything. Auditing and optimising at once produces an account with fourteen simultaneous changes and no way of telling which one helped.

Set a lookback window long enough to contain meaningful volume. Google’s own guidance for judging Target CPA performance is to measure over periods with at least 30 conversions, which is a fair rule for the whole audit: if a segment contains three conversions, you are looking at noise and should say so rather than drawing a conclusion.

1. Is the conversion tracking telling the truth?

Start here, always. The commonest cause of an account that “does not work” is an account that cannot tell what working looks like.

Check each of the following, in the conversions table:

• What is actually being counted. Open every conversion action and find the trigger. A thank-you page view is fine. A click on a button that sometimes fails validation is not. A page view of the contact page is a disaster, and it is remarkably common.
• Which actions are primary. Only the actions you want bidding to optimise towards should be. Newsletter signups, PDF downloads and phone-number clicks sitting alongside genuine enquiries pull Smart Bidding towards the cheapest of them.
• The counting setting. Google lets you count “every” conversion or “one” per ad click, and defaults differ by source — website actions default to every, calls and app installs to one. Lead generation almost always wants one; e-commerce wants every.
• Duplication. A conversion imported from Analytics and tracked by the Google Ads tag is counted twice. Look for two actions with suspiciously similar totals.
• Conversion windows and attribution model. A 90-day window on a business with a two-day sales cycle inflates recent performance and delays the truth.
• Enhanced conversions. These send hashed first-party data — normalised, then SHA256-hashed before transmission — to recover conversions browser restrictions would otherwise lose. If they are off, your conversion count is understated, which affects bidding as much as reporting.

The wider version of this is a measurement problem rather than an ads problem: consent handling, tag firing order and server-side collection decide what the platform ever sees. If your conversion counts changed sharply and nobody deployed anything, read our note on consent mode and server-side tracking before touching a bid.

The uncomfortable finding this step usually produces: cost per conversion is not cost per customer, and nobody has ever compared the two. Put monthly conversions next to the CRM’s closed deals. If the shapes do not match, everything else in this audit is provisional.

2. Read the search terms report properly

The search terms report shows the queries that triggered your ads, matched to the keyword that caught them. Most people skim it for obvious rubbish. That finds a fraction of what is there.

Read it four ways:

1. By spend, descending, no conversions. The direct waste list. Anything with meaningful spend and zero conversions over a period containing enough volume to judge is a candidate negative.
2. By conversions, descending, not in your keyword list. The opportunity list. Terms converting well through broad or phrase match that deserve their own exact-match keyword and their own ad copy.
3. By intent category. Group terms into research, comparison, job-seeking, DIY, competitor, and buying. Most accounts discover they are funding a lot of “how to do it yourself” and “jobs” traffic.
4. By what is missing. This is the one people skip. Google reports only search terms that meet its privacy thresholds, which means a portion of your spend is attributable to queries you will never see. You cannot audit what is not shown, so account structure and match type discipline have to do the work instead.

3. Negative keywords: lists, levels and the ones everyone forgets

Two facts from Google’s documentation change how you should build negatives, and both surprise people who have been running accounts for years.

Negative keywords do not match close variants. A positive keyword will catch misspellings, plurals and close variants. A negative keyword will not. If you exclude “free”, you have not excluded “frees” or a misspelling of it. This is why negative lists have to include plural forms and common misspellings explicitly.

Negative match types behave differently from positive ones. Negative broad blocks searches containing all the terms in any order. Negative phrase blocks the terms in sequence. Negative exact blocks only that exact query with no additional words. Using negative exact where you meant negative phrase is the quiet reason a term you “already excluded” keeps appearing.

Then the structural questions. Are negatives applied at the right level — account-level lists for universal exclusions such as jobs, free and DIY; campaign-level for category separation; ad-group-level to control which of your own ad groups wins which query? Is anything being blocked that should not be, because over-zealous negatives are waste in the opposite direction? And do the limits bite: Google notes that for Display and Video campaigns a maximum of 1,000 negative keywords is considered at account level, which matters for large legacy lists.

4. Smart Bidding needs conversion volume you may not have

The most common structural error in small and mid-sized accounts is a bid strategy the account cannot feed.

Google’s published thresholds are the place to start. Target CPA can be started with no conversion history, and Google says it is effective for campaigns of all sizes. Target ROAS is different: for Search and Shopping campaigns Google states you need at least 15 conversions in the past 30 days at the conversion tracking level, with separate thresholds for Display, App, Demand Gen and Hotel campaigns.

Symptom in the account	Likely cause	Sensible response
Target ROAS on a campaign with single-digit monthly conversions	Strategy chosen aspirationally	Move to Maximise conversions or Target CPA until volume exists
Target CPA set far below the historical average CPA	Target set from a budget wish, not from data	Raise towards the achievable figure and step down gradually
Performance collapsed after a target change	Strategy re-entering a learning period	Change targets in increments, not in leaps, and wait for volume
Six campaigns each with two conversions a month	Fragmented structure starving the algorithm	Consolidate campaigns so the signal is in one place

The point people miss is that consolidation is often the highest-value change available and the least popular, because granular structures feel controllable. Taking the same conversions and putting them behind one bid strategy rather than splitting them across six gives the algorithm one dataset instead of six sparse ones. That is arithmetic rather than ideology, and it is why the fix for “Smart Bidding is not working” is frequently a structural one.

5. Match types, overlap and keywords competing with each other

Look for the same query being eligible in multiple ad groups. The auction resolves the conflict for you, but not necessarily in the direction you would have chosen, and it makes your reporting incoherent because performance for one intent is spread across three places.

Three checks:

• Duplicate keywords across ad groups. Google Ads Editor will find these in seconds.
• Broad match without the conditions it requires. Broad match relies on Smart Bidding and conversion data to do query selection well. On an account with thin conversion volume and no negative discipline, it becomes an expensive discovery tool.
• Brand and generic in the same campaign. This is the single most common reason an account looks better than it is. Brand terms are cheap and convert at a high rate; blended into a campaign average, they conceal generic performance that would otherwise be obviously unprofitable. Separate them, always, and evaluate them separately.

6. Geography, schedule and device: the quiet leaks

These settings are set once at launch and never revisited, which is exactly why they leak.

Location targeting. Google’s default advanced location option targets both physical presence and location of interest — people searching about your area from anywhere. For a business serving one region, that default routinely funds clicks from the other side of the world. Change it to presence where appropriate, and note Google’s own caveat that location is inferred from IP, device signals and search behaviour, and that “100% accuracy is not guaranteed in every situation.” Check excluded locations too; an exclusion added during a supply problem in 2023 is still there.

Ad schedule. Segment by hour and day. The finding is usually not “turn off nights” but “conversion rate collapses when the phone line is unstaffed and the only call to action is a phone number.”

Device. Do not apply a blanket mobile bid adjustment because “mobile does not convert”. Check whether mobile conversion rate is genuinely worse, whether mobile conversions are merely harder to attribute, and whether the mobile landing page is slow — a fixable cause rather than a fact of nature. Our piece on Core Web Vitals covers what to measure.

7. Performance Max: what you can see, and what to do anyway

Performance Max is the campaign type that generates the most argument, and the argument is legitimate: it works well in some accounts and obscures a great deal in all of them.

What you can now see: Google provides a search terms report for Performance Max, with data from March 2023 onwards, segmented by ad format and landing page, though store visits and store sales conversions are excluded. That is more visibility than the campaign type launched with, and it is the first place to look. What remains hard to see is the split of spend across the individual channels inside the campaign, and which assets drove which outcome.

Given that constraint:

1. Exclude brand terms at account level if you want to know what Performance Max earns beyond traffic you already own.
2. Check the asset groups. One asset group covering everything means one audience signal and one creative set for products with different buyers. Split by margin or category, not by whim.
3. Check the product feed if there is one — for retail accounts the feed does most of the work, and missing attributes are usually the actual problem.
4. Look at the placements report for where display and video spend went. The list of apps is often instructive.
5. Judge it on incremental outcomes. New customer acquisition, lead quality, or a holdout test — not blended cost per acquisition, which brand traffic flatters.

Anyone who tells you Performance Max is simply good, or simply a black box, is selling a position. It trades control for reach, and whether that trade is worth it depends on how much of your demand already exists.

8. Landing page and offer mismatch

A meaningful share of “the ads do not convert” is “the page does not convert.”

Click the top ten ads by spend and land where a customer lands. Ask:

• Does the page repeat the promise the ad made, in the ad’s own words?
• Is the specific service in the ad the first thing on the page, or is this the generic services page?
• How long does the page take to show content on a mid-range phone on a mobile connection?
• Is the form asking for more than is needed for a first conversation?
• Is there anything on the page that a buyer in this category would obviously want and cannot find — coverage area, credentials, what happens next?

Sending every ad to the homepage is the single most expensive design decision in paid search, and it is a web design problem rather than a media problem. Fixing it usually beats any bid change in the same account.

9. Creative: assets, pinning and Ad Strength

Responsive search ads accept up to 15 headlines and 4 descriptions, with a minimum of 3 headlines and 2 descriptions. Most accounts submit the minimum, or submit fifteen variations of the same sentence.

Audit for:

• Genuine variety. Headlines should cover different angles — the offer, the proof, the objection, the differentiator, the call to action — not fifteen rewordings of the service name.
• Pinning. Google says pinning is not recommended for most advertisers and can affect Ad Strength, but is appropriate for text that must appear in every ad such as a legal disclaimer, and suggests pinning two or three unique assets to a position rather than one. An account with everything pinned has switched the system off while paying for it.
• Ad Strength, in proportion. Google reports that advertisers improving Ad Strength from “Poor” to “Excellent” see 15% more clicks and conversions on average. That is Google’s own aggregate figure, describing a correlation across many accounts rather than a guarantee in yours — an “Excellent” rating earned by adding weak headlines is not an improvement. Treat it as a completeness check, not a score to game.
• Assets and extensions. Sitelinks, callouts, structured snippets, images, lead forms. Missing assets are unclaimed screen space, and screen space is most of what you are buying.

10. Budget pacing and the shape of your spend curve

Google applies two limits to an average daily budget: a daily spending limit of twice the average daily budget for most campaigns, and a monthly spending limit of 30.4 times it — 30.4 being the average number of days in a month. Knowing this prevents two false alarms and one real one.

The false alarms are a day that spent double the budget, which is expected behaviour, and a month that under-spent, which usually means the campaign was not eligible rather than that budgets were wrong.

The real finding is a campaign that is budget-limited every day. Check impression share lost to budget, and separately impression share lost to rank. Impression share is simply impressions divided by total eligible impressions, so a campaign losing share to budget is one where more money would buy more of something you have already proved works — assuming step one of this audit says the conversions are real. A campaign losing share to rank has a relevance or bid problem instead, and pouring budget in will not fix it.

11. Quality Score, and what it is actually for

There is a persistent belief that Quality Score is a number you optimise directly. Google’s documentation says otherwise: “Quality Score is not an input in the ad auction. It’s a diagnostic tool to identify how ads that show for certain keywords affect the user experience.”

Use it as intended. Its three components — expected click-through rate, ad relevance and landing page experience — tell you where a keyword is weak. A keyword with poor ad relevance needs an ad group whose copy matches it. A keyword with poor landing page experience needs a different destination. A keyword with poor expected click-through rate is probably matched to intent you do not serve.

Chasing the composite number is a waste of an afternoon. Reading the components is a useful hour.

12. The things in the account that should not exist

Every inherited account contains sediment: campaigns for products that no longer exist, audience lists that stopped populating when a tag changed, tracking templates with parameters nothing reads, automated rules nobody can explain still running, auto-apply recommendations quietly making changes you did not authorise, shared budgets that make campaign-level analysis impossible, and disapproved ads nobody has looked at.

None of it is glamorous, and it is frequently where a percentage point of the budget is sitting.

Turning the findings into an action plan

Rank everything you found on two axes — expected impact, and effort — then work in this order.

1. Fix measurement first. Nothing else can be evaluated until conversion tracking is trustworthy, and it is often a small technical job with an outsized effect. Where it becomes a genuine data question — matching spend to closed revenue rather than to form fills — it belongs with your data and analytics work rather than the ads interface.
2. Stop the direct waste. Negatives, geography, excluded placements, campaigns for dead products. Same-day changes, immediate savings, low risk.
3. Fix structure. Brand separation, consolidation for bidding volume, duplicate keyword resolution. Higher impact, more disruption, and each change needs a settling period.
4. Improve the assets. Ads, extensions, landing pages, feed quality. Slower, and the compounding kind.
5. Then, and only then, touch the targets. Bid changes made before the first four steps sit on top of a broken measurement layer.

Change one meaningful thing at a time where the account has the volume to support that discipline, and use experiments where it does not. An account where six things changed on the same Tuesday teaches you nothing.

When to do this yourself

Most of this checklist is genuinely self-serviceable. An in-house marketer with account access and a free morning will find real money in steps one, two, three and six. If your account spends modestly, is built around one service, and someone internal enjoys the detail, that is a perfectly good arrangement — and better than a retainer with a partner who changes nothing between reports.

Outside help earns its cost in narrower circumstances than the market suggests: when spend is large enough that a percentage point matters more than the fee, when the measurement layer needs engineering rather than configuration, when several markets multiply the same fifteen decisions, or when nobody internally has time to look weekly and the account is drifting. That is what our PPC management work is for, and it sits alongside the rest of a digital marketing programme rather than replacing it.

The honest test of any auditor, us included: ask what they would tell you to stop spending on. An audit that finds only things to add is not an audit.