What this audit is for
A full agency audit takes days and produces a deck. This one takes half an hour and produces a short list of things to fix, in order. It is what you run before you decide whether the account needs the deck.
It assumes an assistant connected to the account, because every checkpoint needs live data — ChatGPT or Claude, about five minutes to set up. You can run the same seven checks by hand in the interface; it takes closer to three hours.
Work through them in order. The order matters: the first checkpoint can invalidate the other six.
If you would rather have the assistant do the first pass, the free Google Ads waste audit runs the money-losing half of this list from one prompt and proposes the fixes for your approval.
1. Conversion tracking sanity (5 minutes)
Everything downstream of a wrong conversion definition is efficiently wrong. Start here or do not bother.
Prompt: "List every conversion action in this account with its conversion value, count over the last 30 days, attribution model, and whether it is set as a primary or secondary action."
What signals a problem:
- Two actions with wildly different business value carrying the same or no value — a newsletter signup and a purchase both worth 1, or both worth nothing.
- A primary action that fires on a page view rather than a real event.
- Conversions counted "every" rather than "one" for a purchase, or the reverse for a lead.
- Zero conversions recorded in 30 days on an action the bidding relies on.
Why it matters: Smart Bidding optimises toward the values you declared. If your values are wrong, the platform will spend your budget getting you more of the wrong thing, faster than a human ever could.
Fix first if: anything here is wrong. Nothing else on this list is worth doing until it is.
2. Wasted search terms (5 minutes)
The most reliable money leak in paid search, and the easiest to quantify.
Prompt: "List search terms from the last 60 days with spend above €20 and zero conversions. Show total spend for that list, and what percentage of account spend it represents."
What signals a problem:
- More than 15–20% of spend on non-converting queries in a mature account.
- Terms that are clearly a different product, audience, or country.
- Job-seeker, DIY, free, and competitor-name queries you never intended to buy.
Careful: not all expensive non-converting traffic is waste. Upper-funnel research queries can be exactly right for a considered purchase. This is a judgement call and it is yours — ask the assistant to explain why it thinks each one is irrelevant, then disagree where it is wrong.
Fix: add the ones you agree with as negatives, at campaign or list level.
3. Budget pacing and spend concentration (4 minutes)
Prompt: "For the last 30 days, show each campaign's daily budget, actual average daily spend, and whether it was limited by budget. Then show what share of total spend each campaign took."
What signals a problem:
- A campaign consistently spending at 100% of budget while converting well — you are capping your best performer.
- A campaign spending under 50% of its budget — the budget is not the constraint; something else is.
- One campaign taking more than half of account spend without matching results.
- Budget-limited campaigns and underspending campaigns coexisting. That is money in the wrong place, and it is the single most common easy win.
4. Bidding strategy fit (3 minutes)
Prompt: "For each campaign, show the bidding strategy and the number of conversions in the last 30 days. Tell me plainly which campaigns do not have enough conversion volume for their strategy."
What signals a problem:
- Target CPA or Target ROAS on a campaign with fewer than ~30 conversions a month. The algorithm is optimising noise.
- Manual CPC on a high-volume campaign — you are hand-doing the one job automation is unambiguously better at.
- A target that has never been hit, set months ago and never revisited.
- Maximize Conversions on an account whose conversions are unequal in value, where Maximize Conversion Value was the intent.
5. Keyword and ad group sprawl (4 minutes)
Prompt: "Show ad groups with more than 20 keywords, and ad groups with fewer than 3. For the large ones, tell me whether the keywords are one theme or several."
What signals a problem:
- Ad groups with 50+ loosely related keywords. Nothing in that group can have message-matched ad copy.
- Dozens of single-keyword ad groups with two conversions between them — structure for its own sake, splitting data too thin to bid on.
- The same keyword in several ad groups competing with itself.
- Keywords with hundreds of impressions and zero clicks over 90 days, sitting there diluting relevance.
6. Creative coverage (4 minutes)
Prompt: "For each ad group, show how many responsive search ads exist, how many distinct headlines each has, and the ad strength. Flag ad groups with only one ad."
What signals a problem:
- Ad groups with a single ad and no variant. Nothing is being learned.
- Fewer than 8–10 genuinely distinct headlines. Count *angles*, not headlines: five rewordings of "best price" is one angle.
- Ad strength "Poor" across the account, which usually means missing assets rather than bad writing.
- No sitelinks, callouts, or structured snippets — free real estate left on the table.
Then ask: "Read my landing page at [URL] and tell me which promises in these ads the page does not support." Message-match failure is a top cause of a good click-through rate with a bad conversion rate.
7. Geo and schedule leakage (3 minutes)
The checkpoint everybody skips, and it is often the fastest fix in the account.
Prompt: "Break down spend and conversions by location and by hour of day for the last 60 days. Flag any location or time block with meaningful spend and no conversions."
What signals a problem:
- Spend in countries or regions you do not serve. Usually a targeting setting including "people interested in your location" rather than people in it.
- Overnight spend on a business that only converts during staffed hours.
- A city taking 20% of spend at triple the account's cost per conversion.
Reading the results
Once through, you will have a list. Order it this way:
1. Anything broken in checkpoint 1. Tracking errors make every other number a lie. 2. Geo and schedule leakage. Usually the fastest saving per minute of work. 3. Negatives from checkpoint 2. Recurring saving, low risk — which match type and which level decides whether it stays low risk. 4. Budget reallocation from checkpoint 3. Free performance from money already committed. 5. Bidding mismatches from checkpoint 4. Real upside, needs a fortnight to judge. 6. Structure and creative, 5 and 6. Highest effort, slowest payback — genuinely important, but not first.
What to hand back to a human
This audit finds the mechanical problems, which is most of what is wrong with most accounts. It does not answer the questions that decide whether the account should exist in its current shape:
- Are we bidding on the right intent at all, or buying traffic that was never going to convert?
- Is our offer competitive against what the other four advertisers on this query are promising?
- What is a conversion actually worth to the business, including repeat purchase?
- Should this budget be in search at all?
Those are strategy, and strategy is not in the ad account. An audit tells you the plumbing is sound. It does not tell you the house is in the right place.
Running it as a routine
The value of this list is repetition. Checkpoints 2, 3 and 7 are worth running monthly — twelve minutes — and the full seven once a quarter. Save the prompts. Between quarters, a ten-minute weekly review catches the things that break fast: a dead conversion tag, a capped budget, a fresh disapproval.
Nothing in checkpoints 1 through 7 changes your account: every prompt here reads. The fixes are separate, deliberate steps, and anything that moves money should be proposed for your approval before it applies rather than executed because a chat inferred you wanted it — see the documentation for how that gating works.