The short answer
They will never match, and that is expected. The useful exercise is to explain the gap once, so you know which direction it runs and roughly how large it is, then stop trying to reconcile it monthly.
Run the business on your own enquiry count. Use platform numbers to optimise campaigns.
The five causes
- Attribution window: the platform counts a conversion weeks after the click, in the period of the click rather than the enquiry.
- Blocked tracking: some conversions never reach the platform, so it undercounts.
- Duplicate counting: browser and server events both firing without deduplication.
- Other channels: phone and email enquiries the platform cannot see.
- Definition: the platform counts a form submission, your CRM counts a qualified lead.
Causes one and two push in opposite directions, which is why the gap can look small while both effects are large.
Quantify each once
| Cause | How to size it |
|---|---|
| Attribution window | Compare by click date and by enquiry date |
| Blocked tracking | Compare enquiries received against conversions reported |
| Duplicates | Check whether deduplication identifiers are set |
| Other channels | Count phone and email enquiries separately |
| Definition | Compare raw submissions against qualified leads |
Doing this once gives you a rule of thumb for the gap. After that, a gap outside the usual range is a signal that something broke rather than a monthly puzzle.
Which number to use for what
Platform conversions are for optimising campaigns, because that is what the platform's algorithm consumes. Your own enquiry count is for deciding whether advertising is working.
Mixing them produces the common argument where marketing reports conversions and finance reports enquiries and neither believes the other.
Fix the fixable ones
Duplicates and blocked tracking are technical and fixable. Attribution windows and channel gaps are inherent and need explaining rather than solving.
Start with duplicates, because double counting inflates results and misleads optimisation in a way undercounting does not.