How a Google Ads conversion tracking audit works
Most Google Ads audits start inside the account. This one starts after the click.
Everything from the landing page onward happens on your own property, and it is usually the part nobody independently checks. This page sets out exactly what that check involves, and, at greater length, what it cannot establish.
The chain between a click and the money
The rule marks the handover. Campaign audits grade what sits to its left. This review grades the evidence produced to its right, which is what every decision on the left is made from.
Four steps, and a person at the end of them
- 01
Your site is used the way a paid visitor would use it
Arriving from a paid click, moving through the page, and reaching the point where an enquiry would be made. Nothing is installed and nothing of yours is changed. The check only observes what your site already shows every visitor.
- 02
What the site does, and when, is recorded
Which requests are sent and at what moment. Whether the identifier that ties a visit back to the ad survives the journey. Whether analytics and tag infrastructure are present. How the form behaves when it is used. The order of events matters as much as their presence: a conversion request sent before any meaningful action is a different finding from the same request sent after one.
- 03
Each observation is labelled by the evidence behind it
What the check saw, what it means in plain terms, why it matters commercially, and what would need account access to settle. The label is not decoration. It is the difference between something we can stand behind and something we would be guessing at.
- 04
A person reviews it before it is sent
Our diagnostic technology performs the initial checks, but findings are reviewed before they're sent so that publicly observable evidence isn't presented as something it cannot prove. The judgement about what an observation might mean commercially is not automated, and the report carries a named person's accountability for it.
Why every finding says how confident it is
Behaviour seen publicly and recorded exactly as found, including the timings. A fact, stated without qualification, and reproducible by anyone who repeats the same steps.
The same observation, restated so it can be understood without knowing anything technical. If a finding cannot survive that translation, it is not ready to be sent, because a report only a specialist can read cannot be acted on by the person paying for the advertising.
What the observation could mean for the business, always conditional. Note the modal verbs throughout the reports: may, could, if. They are not hedging, they are the accurate tense for evidence gathered from outside an account.
The specific question that cannot responsibly be settled from outside, so no answer is offered. The badge has no fill on purpose: the absence of fill is the absence of knowledge. It is never styled as an error, because not knowing something is not a fault.
Two badges sit at the top of each finding rather than inside it. Observed externally means every part of it came from public evidence. Observed, with sound elements marks a finding that records something working correctly alongside something worth checking, which is the honest result more often than most audit pages suggest.
What this check cannot do
Published rather than buried, because an audit tool that implies it can see everything from outside is the thing this one is defined against.
- It cannot tell you a conversion action is misconfigured. It can tell you a conversion request was sent at a moment that deserves explaining. Whether that action is Primary, Secondary, or bid toward at all is inside the account.
- It cannot quantify anything in money. No figure for wasted spend, no percentage recovered, no projection. Any such number would require the account, the margins, and the sales data. Any audit that offers one without them is inventing it.
- It cannot see server-side measurement. If conversions are sent from your server rather than the browser, that traffic is invisible from outside. Absence of an observed request is not evidence that measurement is missing.
- It cannot reconcile conversions against real sales. Whether the numbers in your reports correspond to customers who actually paid you is a question only your own records can answer.
- It cannot confirm what your CRM received. A form that appears to submit correctly may still fail on the way to wherever the lead was supposed to land.
- It is a snapshot, not monitoring. A site checked today may behave differently tomorrow, particularly if tags are managed by more than one person.
Public checks can identify what deserves investigation. An authorised audit lets us inspect the actual Google Ads, GA4, Google Tag Manager and conversion configuration behind those findings, which is where most of the list above stops being unanswerable.
See what a finding looks like before you ask for one.
What can you determine without account access?
Whether paid-click attribution reaches your site and survives the journey through it. Whether analytics and tag infrastructure are present. When conversion requests are sent, and after which actions. How your enquiry forms behave. Whether phone enquiries appear to be measured at all. And how the landing experience performs on a mobile connection.
All of it is behaviour any visitor could observe. None of it requires a login.
What can't you determine without access?
Whether a conversion action is configured as Primary or Secondary. What your campaigns are actually bidding toward. Whether server-side measurement exists behind the scenes. How reported conversions reconcile against real sales. What your CRM received, and when.
These are the questions that decide whether an observation is harmless or expensive, which is why they are named in the report rather than quietly left out. The list of what could not be determined is usually the most useful section of any check we send.
How can conversion measurement be checked from outside an account?
A website tells you a great deal about its own measurement simply by being used. Arriving from a paid click, moving through the site and submitting an enquiry produces observable behaviour: requests are sent, identifiers are either carried forward or dropped, and events fire at particular moments rather than others.
Recording what happens and when, especially the order of things, is enough to identify where measurement deserves a closer look. It is not enough to conclude that anything is wrong, which is why findings are labelled rather than asserted.
Is this automated, or does a person review it?
Both, in that order. Diagnostic technology performs the initial checks, because a machine is better at noticing the exact moment a request was sent than a person watching a page.
Every finding is then reviewed by Jose Tabbada before anything is sent, so that publicly observable evidence is not presented as something it cannot prove. The judgement about what an observation might mean is not automated.
Will the check affect my website or my tracking?
No. The check observes your site the way an ordinary visitor does. Nothing is installed, nothing is changed, and no configuration of yours is touched.
One honest caveat: because the check involves visiting your site, the visit itself may appear in your own analytics, in the same way any other visit would.