Series
Conversion tracking
Why your platform numbers disagree with reality, and how to get measurement you can actually make decisions on.
18 articles so far, in the order they were written.
Then, in order
-
Four ways the number gets wrong, and none of them look broken
Why conversion tracking is wrong in most accounts, the four failure modes behind it, and the ten minute check that finds nearly all of them.
Read -
The platform is marking its own homework, and that is fine
Platform reported conversions count what the platform can see and wants to claim. Here is how far they drift from sales, and what to do about it.
Read -
Chrome kept them, and a lot of advice never got the message
Google retired most of Privacy Sandbox in October 2025 and kept third party cookies in Chrome. Here is what that changes, and what it does not.
Read -
Six ways to divide the credit, and why the choice matters less than you think
Attribution models decide which advert gets credit for a sale. Here are the six, what each one distorts, and why the choice is smaller than the argument.
Read -
The model everyone criticises is still the right default for most people
Last click attribution gets blamed for everything. It is biased in a known direction, cheap to run, and better than most of what replaces it.
Read -
The gap is smaller than the internet says, and it should be stable
CRM vs ad platform conversions usually differ by about 10% in my accounts, not the multiples this genre claims. Here is how to size and use yours.
Read -
The server side connection, and the reason most accounts do not have one
The Meta Conversions API recovers events the browser loses. Whether you can run it is usually a question of client access, not of technical merit.
Read -
A score out of ten that nobody will tell you how to pass
Event match quality is scored out of ten and Meta publishes no threshold for a good score. Every number you have been given for it was invented.
Read -
Three ways one action becomes two conversions
Conversion event deduplication stops one action being counted twice. Here are the three causes I find most, and how each one inflates your reporting.
Read -
The uplift figure lives in the marketing, not in the documentation
Google enhanced conversions recover conversions the click ID alone would miss. The published uplift figure is weaker evidence than most people assume.
Read -
Recovering the enquiries your click identifier never matched
Enhanced conversions for leads match enquiries Google could not otherwise connect to a click. Here is what they recover and what the numbers really say.
Read -
Uploading a customer list is not the same as closing the loop
Offline conversion import tells the platform which enquiries became customers. Most accounts that think they do this are running Customer Match instead.
Read -
Worth it on some accounts, overkill on most
Server side tracking recovers events browsers lose. There is no credible adoption figure anywhere, and on most accounts it is not the thing to fix first.
Read -
The threshold that quietly excludes most small advertisers
Google Consent Mode needs 700 ad clicks in seven days per country before modelling starts. Below that you lose the data and get nothing to replace it.
Read -
Estimates arriving in the same column as measurements
Modelled conversions are estimates Google adds where it cannot observe. They sit unlabelled beside real ones and take five days to settle.
Read -
Most people say yes, and the best data comes from an interested party
Consent rates in Europe run from 71% to 89% depending on region. The figures are the best available and come from a consent vendor's own customers.
Read -
Two tools, one week, three reasons they will never match
A GA4 vs Google Ads discrepancy is expected behaviour, not a fault. Three structural differences explain nearly all of it, and none of them are fixable.
Read
Book a free teardown
Send me your worst-performing campaign.
I will tell you what is wrong on the first call. No obligation, no hard sell, and I will say honestly if I am not the right fit.
Book a free 20-minute account teardownReplies within one business day, Lahore time.