Every business running paid media eventually hits the same argument: the ad platform reports forty conversions, the CRM shows twenty-two leads, and sales recognises eleven. Nobody is lying and all three numbers are defensible. They are measuring different things.
Resolving it is less about picking a winner than about knowing which system answers which question.
Why the numbers differ
Ad platforms count conversions against the click that caused them, and they attribute back to the click date rather than the day the conversion happened, which alone makes any day-by-day comparison misleading. They also count multiple conversion actions per session unless configured otherwise.
A CRM counts records created, deduplicated by whatever rule it uses, and it usually has no idea which click produced the record. Sales counts opportunities that passed a qualification bar which frequently exists only in someone’s head.
Which system to trust for what
- Ad platform: relative performance between campaigns, keywords and audiences. It is the only system that can see the click.
- CRM: how many real people entered the pipeline, and what happened to them.
- Finance: revenue. Neither of the other two should be quoted as revenue.
Most reporting disputes dissolve once each system is asked only the question it can actually answer.
The platform is the best instrument for deciding where the next pound goes. It is a poor instrument for telling you what the business earned. Using it for both is what starts the argument.
Aisha Karim, Analytics & CRO Manager, Media @ Marsons
Connect them with the click identifier
The practical fix is to capture the click identifier at form submission as a hidden field, store it on the CRM record, and send outcomes back to the platform through offline conversion import. That closes the loop: the platform learns which clicks became qualified opportunities and which became revenue, rather than which became form fills.
It is a small piece of plumbing and it is the difference between a bidding strategy that improves over quarters and one that becomes very efficient at buying the cheapest possible enquiry.
Agree the definitions before you build
What counts as a lead, what counts as qualified, and at what point a record is created. If sales and marketing disagree on any of the three, the integration will faithfully transmit that disagreement into the bidding algorithm.
Write the definitions down in one place both teams can see. This is the least technical part of the work and it settles more arguments than the integration does.
Watch the deduplication rule
A customer who fills in two forms is one person. Whether your CRM knows that depends on a matching rule somebody configured, and if it is set on email alone it will treat the same company enquiring twice as two unrelated leads. Test the rule against your messiest existing records rather than against clean examples.
Expect the reconciliation to stay imperfect
Even with the loop closed, the numbers will not match exactly, and they do not need to. Cookie limits, cross-device journeys and people who call instead of clicking all put a floor on how close you can get. The goal is a stable, explainable gap rather than a perfect one.
Where the gap changes suddenly, that is the signal worth investigating: it usually means tracking broke rather than that behaviour changed. Our digital marketing hub covers how the measurement layer sits across channels.
Consent changes what can be joined
Where a visitor declines tracking, the click identifier may never be captured, which means those enquiries cannot be tied back to a campaign at all. That is a legitimate outcome rather than a bug, and it puts a permanent floor under how complete the reconciliation can be.
The practical consequence is that platform-reported conversions and CRM records will diverge by a share that varies with your consent rate. Knowing roughly what that share is turns an alarming gap into an expected one, and stops teams chasing a discrepancy that cannot be closed.
Backfill carefully when you first connect
Sending months of historical outcomes into a bidding model in one batch distorts it, because the model treats them as recent signal. Import a short recent window, let the strategy settle, then keep the feed running forward. The temptation to load everything at once is strong and it produces a fortnight of strange bidding.
Two systems, two answers?
We will reconcile your ad platform against your CRM and wire the outcomes back, so bidding optimises toward booked revenue rather than form fills.
