Audits of accounts built by someone else produce a narrower range of findings than you would expect. The same dozen or so problems recur across industries, budgets and agencies, and most of them share a single characteristic: they are invisible from inside Google Ads' own reporting. An account can report thousands of conversions and a conversion value larger than anything the company has ever invoiced, while the sales team can name one deal that came from paid search. Nobody involved was being dishonest. The measurement was wrong at the root, and every report built on top of it inherited the error.

What follows is a list of common cases, in roughly the order we work through them on a takeover.

The thank-you page is doing the measuring

A conversion tag sitting on a thank-you URL fires whenever that URL loads. It fires on a refresh, on a back-button return, on a bookmark, and on anyone who reaches the page directly. In some accounts we have found the same tag misconfigured to count every occurrence rather than one per click, so a single enquiry produces several recorded conversions and the inflation compounds quietly over months.

Every conversion action is marked primary

Newsletter signups, guide downloads, contact form submissions and page loads all feeding the bidding algorithm at once is among the most frequent setups we inherit. Google then optimises toward whichever of those is cheapest to acquire, which is almost never the one that produces a salesperson's phone call. Accounts in this state tend to look efficient on cost per conversion while producing very little pipeline.

Values have been invented

Conversion values attached to actions with no monetary outcome appear regularly, often in round numbers somebody entered once and forgot. In ecommerce, a value per conversion is a reasonable thing to record. In B2B, where a deal takes months, involves procurement and closes at a figure nobody could predict from the click, assigning a value to a form submission invents information the account does not have. Those figures then surface in reporting as revenue, and where value-based bidding is running, they steer spending toward whichever action carries the largest number somebody typed.

One submission fires several tags

Tag containers accumulate. A conversion tag added for a specific form ends up sharing a trigger with a general form-submission rule, and a second language version of the same tag picks up the same trigger. We have seen containers where one submission fired three separate conversion actions plus a fourth on the resulting page load. The account then bids as though it receives four times the enquiries it does.

The CRM signal exists but is excluded from bidding

This one is more frustrating than the others because the work was already done. An offline conversion import sits in the account holding the only data that describes a qualified enquiry, and it has been left out of the goal set or marked secondary. The account cannot learn from the one signal that reflects a real buyer.

Spend is leaving the search results

Search partners and display expansion arrive switched on by default, and on inherited accounts we frequently find both still running inside campaigns named for search. Where it has gone unnoticed for long enough, a substantial share of total spend ends up on display placements: Android APK download sites, VPN utilities, junk-cleaner apps. Keyword-level click-through rates in those accounts look exceptional, which is the symptom rather than the achievement.

Google has been editing the account

Auto-applied recommendations are a setting most people never look at. Left on, it adds and removes keywords on its own initiative, including broad match terms nobody chose. We have reviewed accounts where it accounted for more changes than the person responsible for the account made in the same period, which is how a carefully built keyword set drifts into generic category terms within a year.

Broad match without a negative list built from evidence

Broad match works when the conversion signal is clean and the negative list is maintained from the actual search terms report. It behaves very differently on an account whose conversion signal is counting page loads, because the system optimises confidently toward the wrong outcome. The negative lists we inherit are usually short, generic, and assembled before launch rather than from anything the account has since learned.

Brand and customer traffic is bought by accident

Bidding on your own brand is a defensible decision in a market where competitors bid on it, and plenty of accounts should be doing it deliberately, in a campaign of its own with its own budget. We have seen accounts where generic campaigns pick up brand queries and navigational searches through broad match, charge them as new demand, and mix them into the performance of everything around them. A campaign targeting category terms looks efficient because a share of its conversions came from people already typing the company name.

Existing customers heading to a login page are the clearest version of it. They convert on nothing, they cost whatever the auction charges, and no report separates them from a prospect who has never heard of the company.

The account was left running unattended

Change history tells a story that monthly reporting does not. A common pattern is heavy activity for the first few months after launch, then almost nothing, while the budget continues to spend every day. We have found stretches of the better part of a year with barely a handful of human changes, and the reports issued during those periods rarely mention it.

Budget sits with the generic terms and the enquiries come from the specific ones

When spend is mapped against CRM outcomes rather than platform conversions, the pattern is usually stark. Ad groups targeting broad category terms consume the majority of the budget and produce a handful of contacts between them, while narrowly defined technical, regulatory or product-specific themes generate nearly every enquiry of consequence on a fraction of the spend. Platform reporting makes none of this visible, because it has been counting all of it as conversions.

Measurement is being lost to consent, or will be

Accounts targeting European visitors without a working consent banner are losing measurement across Google's platforms, and the loss does not announce itself. The opposite case creates its own problem: a banner goes live mid-quarter, recorded conversions fall, and the change is read as a performance decline. Both situations need identifying before anyone draws conclusions from a trend line.

Successive owners have left contradictory settings behind

Accounts that have passed through several hands carry the sediment of each one. Two conversion actions track the same event under different names because each agency built its own rather than adopting what was there. Campaign naming follows three incompatible conventions, so any report that groups by campaign name produces nonsense. A structure half migrated to a newer approach sits alongside the older version, both enabled, competing in the same auctions.

Alongside the duplicates sit the things nobody switched off on the way out. Universal Analytics tags still firing years after the platform stopped processing data. Conversion actions pointing at lead form assets or pages that no longer exist. Triggers built for a form that was replaced two redesigns ago, and audience lists nobody has excluded or targeted since the agency that made them left. None of it necessarily causes harm. It does mean the person responsible cannot say with confidence what fires when, and that uncertainty is where the measurable problems have been hiding.

These problems share one thing: none of them shows up in the reporting. An account with inflated numbers looks like an account that is working, and Google will not warn you that a conversion action is counting the wrong event. Someone has to open the account, break the conversions down by action, and check them against the leads the sales team actually received. That is where an audit starts, and it usually changes the headline number the client has been reading all year.