Ads Conversion Audit
What we find

Four real checks, and what each one turned up

These are individual external checks of four businesses that were never clients, carried out using only what their websites show any visitor. Every measurement below was taken. Nothing has been rounded up into a story.

A real finding from an independent external check, anonymised. The business was not a client of Ads Conversion Audit, no account access was used, and nothing is concluded beyond what the check could observe.

None of these businesses was told their website was broken

Each card says what the check saw, what that could mean commercially, and the specific question that would need account access to settle. The second one records a setup that was largely sound, which is the honest result more often than a page like this suggests.

A conversion appeared to be recorded 2.3 seconds after someone arrived

Observed externally
Business type
Consumer business
Region
New Zealand
Access
External check only. No account access.
Basis
Network requests observed during an ordinary page visit
What the check saw
  • A Google Ads conversion request appeared 2.3 seconds after the homepage started loading. Nothing had been submitted, no enquiry had been made, and no meaningful customer action had occurred.
  • The same site also had a separate conversion event visibly connected to its enquiry form.

In plain terms Google received a success signal before the visitor had actually done anything.

Why it matters commercially

The second observation is why this matters. Something was installed and working, so "is conversion tracking set up?" is the wrong question. The question is what is being counted. If an ordinary page visit can produce a conversion, the number in the report stops meaning what most people assume it means.

What we would verify next

What that conversion action represents, whether it is included in the headline performance numbers, whether campaigns are being optimised toward it, and whether it should instead occur after an enquiry, purchase or booking.

Technical detail

A conversion request was observed on the initial page view, ahead of any form submission or interaction event. The public check can confirm the request occurred and when. It cannot see whether that conversion action is set as primary, or whether any campaign uses it as a bidding goal, because both live in the account.

A business asking people to call, with no visible call measurement

Observed, with sound elements
Business type
Specialist healthcare
Region
Australia
Access
External check only. No account access.
Basis
Nine pages inspected, plus a simulated mobile render
What the check saw
  • Across nine pages, the main Google advertising and analytics systems were consistently present, and paid-click information remained available as visitors moved through the site. A lot of this setup looked healthy.
  • The website actively encouraged prospects to phone the business, but the check did not find Google Ads website-call conversion tracking.
  • Under simulated mobile conditions, the main content took 13.1 seconds to render.

In plain terms Phone enquiries may not be getting credit in the advertising reports. And some of the people you have already paid to bring to the site may be leaving before the page finishes appearing.

Why it matters commercially

If a meaningful share of enquiries arrive by phone, those enquiries may not be reaching the advertising reports at all. Campaigns producing real patients would then look weaker than campaigns producing form fills, and the usual response to a campaign that looks weak is to reduce it.

What we would verify next

Whether calls are measured through a route the public check cannot see, whether phone leads are imported from another system, which actions Google is actually using for optimisation, and whether the reported results match what the practice receives.

Technical detail

Google Ads and analytics tags were present and consistent across the sampled pages, and paid-click parameters persisted across navigation. No call conversion mechanism was observed. Call tracking is frequently handled by a third-party provider or imported as an offline conversion, neither of which is publicly visible, which is why this is recorded as a gap to confirm rather than a fault.

A paid visitor could wait more than 20 seconds for the page to appear

Observed externally
Business type
B2B supplier
Region
Europe
Access
External check only. No account access.
Basis
Simulated mobile render, plus attribution persistence across navigation
What the check saw
  • The business was confirmed to be actively advertising.
  • During the mobile test, the primary page content took 20.6 seconds to render. The page scored 47 out of 100 for performance and weighed approximately 4.5 MB.
  • The advertising click reached the site and survived as the visitor moved to another page.
  • One of the browser mechanisms normally used to connect a later customer action back to the original advertising click was not observed during the test.

In plain terms Someone can arrive from an ad, look around, and enquire later. We found something worth checking in the chain that tells Google where that customer originally came from.

Why it matters commercially

The two observations point in different directions, which is why neither is a verdict. Attribution partly worked. The page speed means some people who were paid for may never have seen the page. And in a business where an enquiry often comes days after the first visit, the chain connecting that enquiry back to the click is the part worth checking.

What we would verify next

Whether another method handles that connection: server-side measurement, a CRM, or imported offline conversions. Any of them would make the missing browser-side mechanism irrelevant.

Technical detail

Largest contentful paint measured 20.6 seconds under throttled mobile conditions, with a total transfer weight of roughly 4.5 MB. Click identifiers persisted across an internal navigation, but one of the client-side storage mechanisms typically used for durable attribution was absent. Absence of a browser-side mechanism is not evidence that attribution is broken, because server-side and offline routes are invisible to an external check.

The most dramatic number is not always the most important finding

Observed externally
Business type
Home services
Region
Australia
Access
External check only. No account access.
Basis
Simulated mobile render, plus tag and attribution inspection
What the check saw
  • The business was actively running Google Ads, and paid-click attribution looked reasonably healthy: the click reached the website, survived navigation, and the necessary Google click information was being stored.
  • Google Analytics and a Google Ads tag were both installed.
  • Under the mobile audit conditions, the main content took 52.3 seconds to render. The page weighed 11 MB and scored 36 out of 100 for performance.
  • The site appeared to use external customer-management or lead infrastructure.

In plain terms You can have sophisticated advertising, tracking and customer software. If someone clicks your ad and waits tens of seconds for the page, the leak happened before the lead had a chance to exist.

Why it matters commercially

On paper the measurement setup was in reasonable shape. But the expensive problem here happens earlier than measurement: a person clicks a paid advertisement and waits. Whatever share of those visitors leave first were paid for and never arrived.

What we would verify next

What the real-world load time looks like for actual paid visitors rather than a simulated device, and where enquiries ultimately land, given the external lead infrastructure.

Technical detail

Largest contentful paint measured 52.3 seconds under throttled mobile conditions, with an 11 MB transfer weight and a Lighthouse performance score of 36. Attribution parameters and click identifier storage were both present. The presence of external lead infrastructure is noted because it reinforces the operating rule of an external check: it must not assume where an enquiry ultimately goes.

Two worked examples

Why a conversion count on its own tells you very little

The two examples below are hypothetical, and they are not findings. Neither number could be observed from outside an advertising account, because conversion counts and campaign spend are only visible inside it. They are here to show the shape of the problem.

Say the report from your agency shows 47 conversions this month.

The natural assumption is that those are enquiries, bookings or sales. But without knowing what is being counted, how would you know what those 47 things actually are?

They could be 47 site visits, if a page view is being recorded as a conversion.

The number itself tells you nothing until you know which action produces it. That is the question an independent check answers.

Suppose two campaigns each spend A$2,000 in a month.

One produces twelve enquiries through the website form. The other produces fifteen enquiries, but eleven of them arrive by phone.

If calls are not being measured, the second campaign reports four conversions and looks like the weaker one.

Both numbers are real. Only one of them describes the business. This is a hypothetical, used to show the shape of the problem rather than anything observed on a particular account.

None of this tells you anything about your own site.

These are four other people's setups. The only way to know what yours does is to check it.

Request a free checkHow the check works →
Questions about these examples
Are these real businesses?

The observations are real and the businesses are real, but every identifying detail has been removed and none of them are presented as clients.

Some of them combine observations from more than one check rather than describing a single engagement. That makes them composites, and none of them represent a client relationship.

Do you ever name the businesses you check?

No. Nothing identifying a business we have checked is published, screenshotted or described specifically enough to work out, and that holds whether or not they ever become a client.

It applies to businesses that never asked to be checked as much as to those that did.

How common are these findings?

We do not publish a figure, because we do not have one we could defend. Claiming that some percentage of advertisers have a measurement problem would require a study nobody here has run.

What can be said is narrower and more useful: these are the patterns that recur often enough to be worth checking for, which is why the methodology checks for them.

What happens if nothing is wrong?

You get told that, in writing, with the same detail as any other result. Measurement that was checked and looks correct is labelled verified, and a check that concludes as much is a legitimate outcome rather than a failed one.

It is also the more useful result in one respect: it tells you which numbers you can rely on, and narrows what still requires account access to settle.