Almost every in-house audit starts in the wrong place. You open the account, sort by spend, find the expensive campaigns and change the bids. A month later the dashboard looks slightly better and the bank account looks identical, because that was never where the problem was.
A useful audit is a sequence, not a checklist. Each question only makes sense once the previous one has been answered: if conversions are counted wrong, every bidding decision inherits that error amplified. Below are the seven, in the order they have to be done, each with the specific figure to look at.
If you want the first cut done in a minute before you start, export the campaign report and run it through the dead spend auditor: it separates what has produced results from what has not. The file never leaves your browser.
1. Is what the account counts as a conversion what you call a sale?
This is the first question and the one that produces the most findings, which is why it goes before anything else. Open Goals > Conversions and look at three columns almost nobody looks at: which actions are marked as primary (those are the ones steering the algorithm), the counting type and the attribution model.
- Too many primary actions. If "view contact", "add to cart" and "purchase" are all three primary, you are asking Google to optimise for all three equally. A purchase is worth a hundred times more and the system does not know that.
- The wrong counting type. A form counted as "every" instead of "one" multiplies conversions your CRM will never see. It is the most common mistake in lead generation, and it is the reason the platform's cost per action drifts away from what a customer actually costs you — the distinction is in CAC vs CPA.
- Duplicated conversions. The same purchase arriving through the Google Ads tag and through the GA4 import at the same time. It shows up fast: divide conversions by real backend orders and check whether the number looks suspiciously like two.
If you find something here, stop. Do not carry on with the rest until it is fixed: everything that follows leans on this number.
2. How much of your spend has produced no result at all?
Before optimising anything, you need to know how much budget is going down the drain with no room for argument. The question is not "which campaign has a bad ROAS": it is which campaigns, ad groups and search terms have spent money with zero conversions over a long enough window.
Pick a period of 60 to 90 days (30 does not give you the volume to claim anything) and sort by spend with conversions = 0. In most accounts we review that block sits between 10% and 25% of the investment, and part of it is decidable on the spot: enough spend, zero results, nothing to wait for.
One caveat that stops you cutting where you should not: if your order value is high and your sales cycle is long, a campaign with zero conversions in 30 days can be perfectly healthy. The cut is validated against the full history, not against the window you happened to have open. If your cycle runs into months, the figure that tells you what you can afford to pay is the LTV to CAC ratio, not this month's cost per conversion.
3. Does the structure separate things that behave differently?
The question is not "how many campaigns do I have", it is whether things that demand opposite decisions are mixed together. Three mixes to always look for:
- Brand with generic. If they run together, the aggregate ROAS is a comfortable lie: brand inflates the result of a generic campaign that may not stand up on its own. Split them and look again.
- Performance Max eating everyone else. PMax will take brand queries and Shopping traffic if you let it, and then show up as the star campaign. Look at the search terms report of your Search campaigns before and after PMax launched: if they dropped off a cliff, that is not incremental revenue, it is cannibalisation.
- Fragmentation with no purpose. Twenty ad groups with three conversions a month each never learn. If no different decision depends on that split, merge them.
The same test applies one level up: an account whose platform ROAS keeps looking fine while revenue stays flat is usually an attribution overlap between campaigns, and the metric that catches it is the account-wide one — that is the whole point of MER vs ROAS.
4. Do your bids come from your margin or from an inherited history?
This is where almost everyone starts and where you should arrive fourth. The right question is not "is the tCPA too high?" but where that number came from. In most accounts the answer is that it was set months ago by copying the CPA the account happened to be producing at the time, and has been raised every time volume dropped since.
- A target ROAS is calculated from margin, not from history. The real floor is your minimum profitable ROAS: below it, selling more means losing more.
- Careful with uncapped maximisation strategies. "Maximise conversions" with no target CPA spends the entire budget by definition, whatever the cost per conversion turns out to be. It is a legitimate strategy while learning and a problem if it has been in place for months.
- Changes that are too frequent. If bids are adjusted every few days, the campaign lives in permanent learning and nothing you measure means anything.
5. Which queries are actually charging you?
The search terms report is where the most loose money shows up, and also the most uncomfortable one to read because Google discloses less every year. Even so, 90 days gives you plenty of material.
Look for three things, in this order: terms with spend and no conversion (negative keyword candidates), terms with a different intent (informational, jobs, "free", competitors when you do not want them) and overlaps between your own campaigns, which make you bid against yourself and push your own CPC up.
Write down the share of spend that does not appear in the report at all. If more than half of your Search spend is invisible, any conclusion you draw from it is partial, and it is better to say so out loud than to treat it as complete. And while you are there, check whether a cheap CPC is actually good news: CPC vs CPM explains when a falling CPC is a sign of worse placements rather than better auctions.
6. Is the catalogue or the creative holding everything else back?
In ecommerce this question is usually worth more than the bids. Open Merchant Center and look at disapproved products and active products with no impressions: the second group is the silent one, it throws no error and it never competes. Why that happens is in the product feed in Google Shopping.
In lead generation the equivalent is the creative block: ads with fewer than three headlines being used, assets left unconfigured, or a landing page answering a different promise from the ad's. Compare the ad copy against the H1 of the destination page, literally, side by side. The disconnect between the two is one of the most frequent causes of an expensive CPC with poor conversion, and it shows up first as a CTR below the benchmark for its channel.
7. Do the account's numbers match the business's?
The last question is the one that turns the audit into a decision. Take a closed month and put three figures next to each other: Google Ads conversions, GA4 conversions and real backend orders. They will not match, and they do not have to: they measure different objects. What matters is that the gap is stable and known, because then you can correct for it.
Calculate two ratios and write them down: platform conversions divided by backend orders, and GA4 conversions divided by backend orders. Those two numbers are your account's normal. If the gap moves from one month to the next without you having touched campaigns, go back to question 1: the problem is measurement, and the whole audit has to be redone once it is fixed. And keep the three systems on the same timezone — half of all unexplained discrepancies are a timezone offset.
When you finish the seven, do not walk out with a list of fifty findings. Walk out with three actions ranked by the money at stake and a date on each. An audit that does not end in an ordered plan is a document, and documents do not change an account.
If you would rather have this done by someone with their hands inside a thousand accounts, that is exactly what we deliver: the same dimensions, with every finding quantified in euros and the plan already prioritised by margin impact. You can request a viability audit of your account at no cost.
Frequently asked questions
How often should you audit a Google Ads account?
A full run of the seven questions twice a year, and always after a structural change: a website migration, a new checkout, the arrival of Performance Max or a change of agency. In between you do not need to repeat the whole thing: watching question 2 (spend with no result) and question 7 (that the ratios against the backend stay put) once a month is enough. Auditing more often than you execute changes only produces documents.
Where do I start if I only have an hour?
With questions 1 and 2, in that order, and nothing else. In an hour you can check which actions are marked as primary and with which counting type, and pull 90 days of spend with zero conversions. Those two answers already tell you whether the rest of the account is optimising against false data and how much money is decidable today. The other five questions cannot be interpreted until those two are answered.
Can I audit the account myself or do I need an agency?
The seven questions are answered with read access to the account, to Merchant Center and to the backend, and none of them needs a paid tool. What does take experience is the judgement: deciding whether a 30% gap against the backend is normal in your sector, or whether a campaign with no conversions in 60 days is dead or simply has a long cycle. Running the diagnosis in-house and pressure-testing the doubtful decisions outside is usually the most efficient split.
What date range should an audit look at?
60 to 90 days for anything that is a cutting decision (spend with no result, search terms, products with no impressions) and one closed month for the reconciliation against the backend. With 30 days there is not enough volume to claim that something does not work, and beyond 120 you are mixing seasonality and different configurations. Always with the same timezone across all three systems.
What is dead spend in an account?
It is the share of the investment that has produced zero results according to the platform itself over a window long enough to say so. It is not the same as a bad ROAS: a bad ROAS still produces sales, dead spend produces nothing at all. It is the cleanest block to cut because it needs no forecast — only the check that the window fits your sales cycle. The dead spend auditor splits your campaign export into exactly that: spend with results, spend with no results, and the part of it that is already decidable today.
How much of your budget produced nothing?
Export your campaign report and drop it into the dead spend auditor: it tells you how much spend produced zero results, how much of that is already decidable, and which pieces to switch off, fix or hold. It runs in your browser: your data never leaves your computer.