Google Ads Audit: Checklist, Costs and When It Pays Off

Last updated 28 September 2026.

Search for “Google Ads audit” and what you mostly get is offers: agencies that will examine your account for free, and a contract at the end. That is no reproach, the model works because there is something to find in almost every account. But it does not answer the question underneath: what is actually being checked, how much of it can I do myself, and what does the rest cost? This article walks through the nine areas of an audit, says for each one where the money typically leaks, and attaches a checklist you can tick off, with no form in front of it.

What a Google Ads audit is, and when one is due

A Google Ads audit is the systematic examination of an ad account for errors, wasted spend and unused potential, from measurement through account structure to search terms, bids and product data. The result is a prioritised list of what should be fixed first. Implementation comes after that.

An audit is due in four situations. When you take over an account, whether from an agency, a predecessor or your own past, because nobody knows what legacy is sitting in there. When results tip over for no obvious reason, the ROAS falls over several weeks or cost per order rises without anything changing in the market. Before expensive periods, so before Black Friday and the Christmas season, when every mistake is paid for with the most expensive traffic of the year. And regularly, because an account that runs three years without a check has certainly accumulated search terms, settings and products that quietly draw budget.

One point belongs right at the start, because it sorts out expectations: an audit finds faults in the account, not faults in the offer. If price, range or landing page do not convince, even the best audit will find nothing that turns revenue around. The ad account is then the wrong place to look.

The usual money burners: what keeps turning up in accounts

After dozens of accounts taken over, the findings repeat so reliably that you can list them in advance. Sorted by the damage they typically do in a shop account, and where we have measured it, the figure is included.

On the data: the wasted-spend and ROAS figures come from e-commerce customer accounts in German-speaking markets, analysed over twelve months to July 2026. The feed and serving figures come from the same accounts in the Merchant Center. The budget figures in the section on budget allocation come from our own daily log of delivery status; the measurement period is stated there.

Search terms with no conversions that still collect clicks month after month. Measured across our customer accounts, around 10 percent of search and shopping budget goes to terms that have never produced an order despite double-digit click counts, and that is in accounts that are already being optimised continuously. In unmanaged accounts the share is higher.

Campaigns stuck at their daily budget without anyone having checked whether they are worth it. In our customer accounts almost every second running campaign was limited by budget at least some of the time, and a good quarter of them practically permanently. And it mostly hits campaigns with below-average ROAS, which turns the usual advice on its head.

Conversion tracking that counts but does not value: every order with the same figure, whether it was 30 or 800 euros. Smart Bidding then optimises for order count instead of revenue, and the measured ROAS is a random number.

Display network and search partners, quietly switched on when the campaign was created, delivering cheap clicks with no purchase intent for online shops.

A single ROAS target across the whole range, so that bestsellers subsidise slow movers and nobody sees which products actually advertise profitably.

A location setting on “presence or interest”, buying clicks from countries the shop does not even ship to.

Products that sit cleanly in the feed and still never get served. In our feed analysis, at the median only a good third of approved products got any paid impressions at all, and for catalogues above ten thousand items only a good tenth. That is not a fault in the account, but it is the biggest blind spot an audit can uncover.

The nine areas that follow are ordered so the biggest levers come first. If you only have an hour, check the first three.

Tracking and conversions: the foundation, and the most common mistake

Without clean measurement every other check is worthless, because everything the account optimises, it optimises towards the numbers it is given. That is why tracking comes first in every serious audit, and why the tracking diagnosis is part of Ginny's first analysis of every newly connected account.

The most common fault is the static order value: conversion tracking reports every order, but all with the same amount. The algorithm then cannot tell a 30-euro order from an 800-euro one and bids the same for both. If you see a suspiciously round or identical value per order under conversions, that is the fault. The second common finding is double counting: an old tag from the shop system and a new one through Tag Manager both fire, and every order counts twice. The ROAS then looks fantastic and Smart Bidding spends money on conversions that only half exist.

On top of that come three checks that have been standard since 2024. Is Consent Mode v2 implemented properly? Without it, European conversions are missing from measurement. Are enhanced conversions active? They restore part of the signal lost to consent. And do only purchases count as the primary conversion for bidding? Add-to-cart clicks or newsletter sign-ups in that role steer the algorithm towards cheap goals.

What to check: is the order value passed dynamically? Does every order count exactly once? Is Consent Mode v2 active and verified? Are enhanced conversions set up? Is the purchase the only primary conversion? Does the number of measured orders roughly match the shop back end?

Search terms and negative keywords: where the budget leaks

The search terms report is where an audit finds money fastest. Google matches ads to queries nobody booked, and some of that is useful while some is pure waste: job searches, how-to queries, “free”, competitor brand names that never buy, spare-part searches for products the shop does not stock.

The check is mechanical: sort the last 90 days of search terms by cost, mark everything with a double-digit click count and zero conversions, and that list is your first exclusion candidate. The benchmark for whether an account is maintained is in the figures above: around 10 percent of budget lands on such terms even in optimised accounts. Anyone at 25 or 30 percent has excluded nothing for months. How to set exclusions without losing revenue in the process: Negative keywords: stop wasted spend without losing revenue.

What to check: when was the search terms report last reviewed? Are there negative keyword lists at account level, and are they assigned to every campaign? What share of cost sits on terms with more than ten clicks and zero conversions? Are brand searches and generic searches steered separately? Are two campaigns bidding against each other on the same terms?

Campaign structure, bidding and budget allocation

Structure decides whether the algorithm can work sensibly at all. The same patterns keep surfacing in audits. Too many keywords per ad group, so one ad is meant to fit dozens of search intents and fits none of them properly. Every product in one campaign with one shared target, so the margin of the individual product plays no part. And campaigns permanently stuck at their daily budget without anyone having decided that they should be.

How often budgets throttle, and which campaigns they hit

Google Ads shows the “Limited by budget” status only as a snapshot, with no history. If you do not log it daily yourself, you cannot tell whether a campaign happened to hit its limit yesterday or has done so every day for weeks. We do log it. We analysed two weeks of daily logs up to 19 September 2026, covering more than a hundred running campaigns in managed online shop accounts. The shares are calculated against the days actually logged, not a rounded month:

More interesting than the frequency is which campaigns it hits. The common assumption, stated in almost every audit guide, is that the budget holds back the good campaigns. In our data that is not the case. Of the permanently limited campaigns, only a good third sat above the ROAS median of their own account; with a purely random distribution you would expect 50 percent. The median ROAS of the permanently limited campaigns was 2.94, against 3.68 for those never limited.

So the daily budget mostly holds back the middle of the field. The actual mistake is that nobody looks at which campaign it hits. For the audit that means: every permanently limited campaign needs a decision. If it sits above its account average, the budget should go up. If it sits below, the cap is the smaller problem. Then the campaign itself needs reworking.

On the method: only Google's explicit budget reason counts as limited, not the combined status that also triggers on low search volume. The ROAS figures are based on the campaign's primary conversions and therefore sit below what the same account shows in its own Google Ads interface. For comparing the two groups that makes no difference, because both are calculated the same way.

With bidding there are two questions. Does the strategy fit the data? A target ROAS needs at least 15 conversions in 30 days according to Google, it becomes stable well above that, and below it swings erratically. We regularly find campaigns running a target ROAS below that threshold when taking accounts over. And does the target fit the margin? A ROAS target below break-even buys revenue at a loss; one far above it throttles delivery into invisibility. How break-even is calculated is in our ROAS guide.

Then the settings that quietly cost money: the location option “presence or interest” instead of “presence”, search partners and display expansion in search campaigns, ad schedules that were never adjusted, and campaigns rebuilt so often in the last 30 days that they never left the learning phase.

What to check: are brand and generic campaigns separated? Do products with different margins have different targets? Which campaigns sit permanently at the budget cap, and are they above or below the account ROAS? Does the bidding strategy fit the number of conversions? Is the ROAS target above break-even? Is the location setting on “presence”? Are search partners and display switched off in search campaigns?

Performance Max and product data: opening the black box

For online shops the bulk of the budget runs through Shopping and Performance Max, and that is exactly where an audit is hardest, because Google shows little. Three things can still be checked, and they are among the most rewarding in the whole audit.

First, Merchant Center status. Disapproved products are not served, and they are not served quietly. The median in our analysis is under one percent, but when it hits, it hits by category, with one outlier at 28 percent of the feed. Price differences between feed and product page are the most common reason, often because of a 24-hour sync window on price changes. The three most common causes and their fixes: Product disapproved in the Merchant Center.

Second, serving. An approved product is allowed to run; whether it runs is decided by Google within the budget. At the median only a good third of approved products get any impressions at all. If the product view of the campaign shows your bestsellers standing still while remainders take budget, you have a prioritisation problem that no error report will flag. The fix lies in custom labels, separate asset groups or a positive selection of the products that should be advertised.

Third, feed quality itself: titles that contain brand, product type and the most important attributes, complete GTINs, current availability. The title is the most important relevance signal Google matches queries against, and in many feeds it still holds the internal item name from the ERP. If you cannot or would rather not check the feed yourself: Google Merchant Center consulting: feed check from €190. The wider context is in our Merchant Center guide.

What to check: how many products are disapproved, and which categories does it affect? What share of approved products received impressions in 30 days? Are the bestsellers running? Do product titles contain brand, type and attributes? Are GTINs complete? Do prices and availability match between feed and shop? Are there separate asset groups in PMax for different ranges?

Ads and landing pages: the area that comes last

This area sits deliberately near the end, because the levers are smaller than with measurement, exclusions and the feed. You should still not skip it, because it is the only part of the account the customer actually gets to see.

Four things can be checked quickly. Whether responsive search ads reach at least “good” ad strength, which in practice usually fails on too few or too similar headlines. Whether the copy names a concrete advantage, so price, shipping terms or range, rather than platitudes like “top quality”. Whether sitelinks, callouts and snippets are in place and still current, because expired promotions in assets often outlive campaigns by months. And whether the landing page shows the advertised product rather than the home page, is reachable, and loads on mobile in under three seconds.

The most common finding in this area is the cheapest to fix: discount assets and promotional copy that expired long ago and are still being served. That costs not a cent extra in click price, but it costs trust with exactly the people who have already clicked.

What to check: do all responsive search ads reach at least “good” strength? Does the copy name a concrete advantage? Are sitelinks, callouts and snippets in place and current? Do final URLs lead to the advertised product? Do landing pages load on mobile in under three seconds? Have expired promotions been removed from ads and assets?

What an audit costs, and how often you need one

The price range is wide, because three different services run under the same word. The free agency audit is a sales instrument: it finds something, because it always finds something, and there is a proposal at the end. That is legitimate, you just want to know that the priority list is sorted towards the proposal. The paid one-off audit in Germany typically runs between €800 and €3,000, with individual providers asking up to €3,500, and for that you get a document with findings and recommendations that you then have to implement yourself. And the ongoing check is part of a service, whether from an agency or software.

On frequency, what keeps proving true in our accounts is this: a one-off audit is a snapshot. Search terms and budget allocation want checking weekly, ads and landing pages monthly, structure and targets quarterly. If you cannot manage that, an overhaul every two years will not help you much. You need somebody or something to take over the weekly items. What clicks, budgets and management cost in total is in our guide to Google Ads costs, and our own prices are on the pricing page.

Check it yourself, agency or AI tool: what fits which shop

Checking it yourself works if somebody on the team understands the account and has the time to go through the checklist at the end of this article quarterly. The weekly items will still be left undone, and that is the honest experience from almost every shop without a PPC role of its own.

An agency is worth it for the one-off audit when the account is large, old and hard to survey and an outside view finds more than your own. For the ongoing check it is a budget question, which we have worked through here: Google Ads agency vs. tools.

An AI tool is the answer to the frequency problem. Ginny checks the recurring items from this article continuously rather than once: search terms with no conversions and campaigns at the budget cap daily, disapproved products in the Merchant Center daily, the Performance Max structure weekly, the tracking on every new connection. What it finds arrives as a proposal with a reason and an estimated effect in euros, and you approve it or turn it down. The audit then runs along in the background instead of landing as a project every two years. Which other tools exist and who they are built for: The best tools to automate Google Ads for e-commerce.

After the audit: prioritise findings and keep at it

The most common fate of an audit is a PDF with forty findings, six of which get implemented. The findings are rarely to blame; what is usually missing is the order. It can be set in three stages.

Measurement first, then everything else. As long as tracking counts wrongly, every further change is a blind flight, and changing bids on the basis of wrong data makes it worse rather than better.

Then the exclusions, because they work immediately and risk nothing: search terms with no conversions, display in search campaigns, locations outside the delivery area. Those are the items that free up budget from day one.

Only after that the structure, and with space around it. Every restructure sends campaigns into a learning phase, so never several at once and never right before a season. And for the items that recur, a fixed rhythm or a system that holds it.

The full checklist to print out and tick off, sorted by priority: download the Google Ads audit checklist (PDF). No form, no email address. It has 43 items rather than the 180 some lists online run to, and that is deliberate: a list that checks everything does not get worked through. One that covers the three most important areas in an hour does.

Frequently asked questions about Google Ads audits