Measurement
Google Ads conversions don’t match analytics: here is why
Published by Zubair Afzal, responsible editor
First published Updated 5 min read
They’re answering three different questions
Say the ad platform reports 41 for the month. Analytics reports 28. The CRM has 19. Somebody concludes that two of the three are broken, and the month is spent arguing instead of deciding.
None of them is broken. An ad platform counts conversions it can credit to a click it served, within a window it chose. Analytics counts events it observed in a session on your site, with whatever identity and consent it had. A CRM counts people somebody typed in. Three populations, three periods, three systems with different information.
The useful goal isn’t agreement. It’s a gap you understand and a decision about which number the business quotes.
The causes
Five mechanisms that produce most of the gap
Work down this list and most discrepancies are explained without opening a support ticket.
- Different questions. One counts credited clicks, one counts observed events, one counts people. A single customer can be one, two or three of those.
- Different dates. Ad platforms book a conversion to the date of the click, so a closed month keeps growing. Analytics and the CRM use the date it happened.
- Different windows. A thirty-day click window and a seven-day window over the same campaign produce different totals from identical behaviour.
- Consent and blocking. Declined consent, blocked scripts and privacy settings mean some systems see an event and others estimate it or miss it entirely.
- Duplicates and identity. A person who submits twice, or who phones after filling in a form, is one prospect in a CRM and two conversions almost everywhere else.
The procedure
Reconcile it in an afternoon, once
The output is a known gap and a chosen number, not matching totals.
Pick one event and one closed month
Not all conversions. One: the main enquiry form or the phone number, whichever produces the most real business. Use a month that closed at least three weeks ago so late crediting has settled. Comparing several events at once is how these exercises stall.
You get: One event, one settled month, three totals written down
Line up the definitions before the numbers
For each system, write down what it counts, the date it uses and the window it applies. Then set the windows as close together as the tools allow. Much of the discrepancy usually disappears here, and what remains is the part worth understanding.
You get: A one-line definition per system, in writing
Name the gap and choose the number you quote
Write the remaining difference down as a ratio, with the reasons beside it. Then decide which system the business quotes for money and which it uses for allocating spend. Put both decisions in the reporting document so nobody reopens them next quarter.
You get: A recorded gap and a single source of record
When it isn’t attribution
The gaps that mean something is broken
Some discrepancies are normal and some are faults. A fault usually announces itself by being unstable: the gap was a third last quarter and is now double, or it changed on a specific date.
The common causes are mechanical. A tag installed twice during a redesign counts every conversion twice. A form firing on submit and again on the thank-you page does the same. A tracking script moved behind a consent banner sees nothing from visitors who never answer it. A campaign link with no tags reports its traffic as direct, putting every enquiry from it in the wrong bucket.
Check the date the gap changed against the date something shipped. That comparison resolves most of these faster than any analysis.
Questions
Common questions about mismatched numbers
Which number should I quote?
Pick the system closest to money and stay with it. For most businesses that’s the CRM or the invoicing system, because it counts people somebody spoke to, not events a browser fired.
Use the platform numbers to decide where to spend, because only they know which click was served. Use the CRM to say what the month was worth. Write down which is which, so nobody reopens it monthly.
Why does last month keep changing?
Because an ad platform credits a conversion to the date of the click, not the date it happened. Somebody who clicked on the 28th and enquired on the 5th is added to the previous month days after it closed.
Modelled conversions do the same for a different reason: estimates are revised as more data arrives. Neither is an error, and both mean a report printed today and one printed in three weeks will differ.
Does consent affect this?
Yes, and it’s one of the larger causes on sites with a consent banner. If a visitor declines, some systems record nothing and others record a modelled estimate, so two tools observing the same visit legitimately produce different totals.
The practical effect is that analytics tends to see fewer events than the server-side record, and the difference varies with how many people decline.
Is it worth paying to fix?
It’s worth paying to make the gap stable and understood. It’s rarely worth paying to make the numbers match, because they measure different things and won’t.
A stable gap is usable. If the platform reliably reports about a third more than the CRM, decisions can be made on either number. An unstable gap is the expensive problem, and usually points at broken tracking, not attribution.
What if the gap is enormous?
Check for double counting first. A form that fires a conversion on submit and again on the thank-you page produces exactly this, as does a tag installed twice during a redesign.
Then check whether the CRM is being filled in at all. A large gap between analytics and a CRM is as often a sales process that doesn’t log enquiries as a tracking fault.
References
Sources for Google ads conversions don’t match analytics
The official documents and research this page is based on. Platform rules change, so check the source before you act on a detail.
Related
Keep reading on Google ads conversions don’t match analytics
Start here
If the gap is unstable, it’s probably a fault
A stable difference between systems is something to document. A difference that moves is usually a tracking problem with a date attached to it. Tell us when it changed and we’ll tell you where to look.
Want to see how your own site stacks up? Get my free SEO audit