Conversion event deduplication is the plumbing that stops one action being counted as two, and when it is missing your reporting inflates quietly rather than breaking loudly. Somebody submits one form. Your dashboard records two leads. Nothing errors, and the campaign that produced them looks twice as good as it is.
The three causes below account for nearly every case I find, and none of them require anybody to have made a mistake.
This is for you if your lead counts look healthy and your sales team cannot find half of them.
What conversion event deduplication is doing
It recognises two reports of one thing and keeps a single count.
The clearest example is Meta running a browser pixel and a server side connection together. A purchase happens, both routes report it, and without something linking the two you would count two purchases. Each event carries a shared identifier, Meta matches them, and one survives.
That mechanism only works where it has been configured. It is not automatic and it is not a setting somebody turns on once for the whole account.
It also only covers that one situation. Two conversion actions independently tracking the same button are not deduplicated by anything, because as far as the platform is concerned you asked for two different things to be measured.
The first cause: two events counting one action
This is the most common by a distance, and it is almost always historical.
Somebody set up conversion tracking two years ago. Somebody else added a tag manager later, configured the same thank-you page, and had no way of knowing the first one existed. Both fire. Both count.
The reason it survives is that neither person did anything wrong, and neither can see the other’s work from where they sit. It takes somebody looking at the whole account to notice.
I find this most often after an agency change, a website rebuild, or the arrival of a new marketing hire. All three involve somebody setting up tracking without a complete picture of what is already there.
The second conversion event deduplication failure: count every instead of count one
This is a single setting, and on lead generation accounts it is usually set wrongly.
Count one records one conversion per click, however many times the action happens. Count every records each occurrence separately.
For an ecommerce purchase, count every is correct, because somebody genuinely buying twice is two sales. For a lead form it is almost never correct. The same person submitting twice, or refreshing a thank-you page, or coming back next week to enquire again, is not several leads.
Call conversions are where this does the most damage. Somebody who rings three times about one job becomes three conversions, and the campaign that found them looks three times more efficient than it was.
Check it per conversion action rather than per account, because these get configured individually and inherit nothing.
The third cause: counting the same lead in three places
This one is subtle enough that people defend it when you point it out.
A site tracks the submit button click, the click on the phone number, and the thank-you page load. All three are set up as conversions. One person who fills in the form, then rings you from the same page, has generated three.
Each of those events is worth measuring. The mistake is counting all three as leads rather than counting one as a lead and treating the others as engagement signals.
The fix is a decision rather than a technical change: name one event as the lead, and demote the rest. Everything else can still be tracked. It just stops being counted in the number the business judges you on.
How to tell duplication apart from ordinary disagreement
Two measurement systems disagreeing is normal. Conversion event deduplication failures look different, and the difference is worth learning.
Ordinary disagreement is untidy. Your platform reports 10% or 15% more than your CRM, the figure drifts month to month, and the causes are spread across spam, timing and integrations.
Duplication is suspiciously neat. It clusters near round multiples, because one action counted twice produces roughly double, and three events counting one lead produce roughly triple.
For scale on what ordinary disagreement looks like, Seer Interactive tested Meta’s incremental attribution across $1.05 million of ad spend on six accounts and found Meta reporting 87% of conversions as incremental where GA4 said 67%. That is a wide gap between two systems both working correctly. It is not a multiple, and that shape is the tell.
So when the ratio sits near 2.0 or 3.0 and holds there, stop looking for subtle causes and go and count your conversion actions.
Why duplication survives longer than the opposite fault
Under counting has a natural enemy. Somebody sees poor performance and goes looking.
Over counting has a protector. The campaign looks efficient, so it gets more budget and less scrutiny at exactly the same time, and the distortion grows in proportion to the money it attracted.
I have seen accounts spend months moving budget toward their best performer, where best meant most duplicated. Every individual decision was correct given the reported figure. The figure was wrong.
That is why conversion event deduplication is worth checking even when nothing looks broken. Nothing looking broken is the symptom.
What happens to your numbers when you fix it
They fall, immediately and visibly, and somebody will ask why.
Prepare for that before you change anything. Your reported conversions drop, your reported cost per lead rises, and none of it reflects any change in performance. The account was always producing what it is now reporting.
Say so in advance rather than afterwards. Explaining a 40% drop before it happens is a competence conversation. Explaining it a week later, after somebody has already escalated, is a different one.
The history does not get corrected either. Your old months stay inflated, so year-on-year comparisons across the fix date are not comparable, and it is worth writing down when it happened.
How to check conversion event deduplication in your own account
Three checks, in order of how quickly they pay off.
List every conversion action and ask what fires each one. If two describe the same moment, you have found it. This is tedious and it is where most duplication turns up.
Then check the count setting on each lead and call action. Anything set to count every occurrence deserves a specific reason to stay that way.
Then, on Meta, open Events Manager and confirm your money event shows a deduplication status rather than two separate counts. Two counts means it is duplicating right now.
What to change this week
Three steps.
Pull your conversion actions into a list with what triggers each one, which usually takes twenty minutes and often ends the investigation on its own.
Then compare one month of platform conversions against your CRM. A gap near a round multiple points at duplication rather than the usual causes.
Then, before fixing anything, tell whoever reads the report that the number is about to drop and why.
What conversion numbers actually measure covers the mechanism underneath, the four ways conversion tracking goes wrong puts duplication alongside the other failure modes, and the Meta Conversions API is where browser and server events have to be reconciled. The case studies show what the numbers look like once they are honest. If you want somebody to audit what your account is really counting, book a teardown.