Every advertising account tells at least three versions of the same week. The ad platform reports one conversion count, the analytics property reports a smaller one, and the sales system reports something smaller again. The Growth Bully, a Malta performance marketing agency, treats most of that gap as normal rather than as a fault, because the three systems are not measuring the same thing and never were.
What is worth fixing is the part of the gap caused by broken plumbing: consent banners that block tags before they fire, browsers that strip identifiers, short cookie lifetimes, and a setup that was correct the day it was built and has not been checked since. That part is recoverable, and recovering it changes which decisions the account makes.
Why do my ad platform and analytics numbers never match?
Because they count different events, over different windows, from different vantage points. The ad platform credits a conversion to the ad that was clicked or seen inside its own attribution window. Analytics credits the last channel before the visit. The sales system counts only what a human confirmed. Three honest systems, three different totals.
Once that is understood, the useful question stops being which one is right and becomes how far apart they are and whether the distance is stable. A platform reporting consistently more conversions than analytics is expected. A platform whose lead over analytics doubled last month without a change in strategy is a signal that something moved in the setup, not in the market.
What is server side tracking and do I need it?
Server side tracking sends the conversion event from your own server to the advertising platform, instead of relying on the visitor's browser to send it. It survives ad blockers, browser privacy restrictions and short cookie lifetimes far better than a browser tag on its own. Any account spending seriously on paid media now needs it.
The important nuance is that server side tracking is not a replacement for the browser tag. The two run together. The browser sends what it can, the server sends the same event with whatever additional identifying information you legitimately hold, and the platform matches the two into one conversion. Setting it up as an either or choice is the most common way this project goes wrong.
What does consent mode actually do to my data?
Consent mode changes what tags are allowed to do based on what the visitor agreed to. Without consent, tags either do not fire at all or send a stripped signal carrying no identifiers. Platforms then model the missing conversions statistically. The reported total still looks continuous, but part of it is now an estimate.
This matters more than most businesses realise, because a modelled conversion cannot be traced back to a person. It counts in the campaign report and it does not appear in your CRM, which widens the gap between the two systems for reasons that have nothing to do with performance. A consent banner redesign can move reported conversions by a visible margin in either direction without a single thing changing in the ads themselves.
How does the Conversions API change what gets counted?
It gives the platform a second, server sent copy of each conversion event, so conversions that the browser could not report still get counted. Sent correctly, each event carries a shared identifier that lets the platform recognise the browser copy and the server copy as one event rather than two. Sent carelessly, it double counts.
Deduplication is the whole game here. An account that suddenly reports far more conversions than the sales system can account for, immediately after a server side implementation, is almost always counting the same conversion twice. That is worth checking before anybody celebrates a step change in reported cost per acquisition.
How do you check whether your tracking is actually broken?
You test it the way a customer would use it, not the way a report describes it. Most broken tracking is not subtle once somebody walks the journey end to end with the reporting open beside them, which is exactly what happens in a pipeline scorecard review.
- Submit a real enquiry yourself from a phone on mobile data, then confirm it appears in the CRM, in analytics and in the ad platform.
- Decline the cookie banner and repeat it, because that is the path a meaningful share of your traffic takes.
- Check the event names for duplicates, near duplicates and legacy events nobody has removed since the last website rebuild.
- Confirm which single event is set as the optimisation target, then confirm it is the one the business actually cares about.
- Compare last month's platform conversions against real enquiries and write the ratio down, so that next month you have something to compare against.
Running that sequence once a quarter catches almost everything. Running it never is how an account spends six months optimising towards a button that stopped firing after a template change.
Which numbers should you actually trust?
Trust the system closest to the money, and use the others for direction. The platform is the best guide to which ad or audience to change, because only it can see the ad. The sales system is the only honest account of what the money produced. Neither replaces the other, and neither should be quietly ignored.
In practice that resolves into a short hierarchy:
- Revenue or closed business is the decision number, and it lives in your CRM or accounting system.
- Qualified enquiries are the operating number, provided everyone agrees on what qualifies.
- Platform conversions are the optimisation number, useful for comparing ads with each other and nothing else.
- Analytics sessions and channel reports are context, best read as trends rather than totals.
The failure mode is mixing hierarchies inside one conversation, where a platform cost per lead is compared against a CRM lead count and somebody concludes the campaign is failing. Two different denominators, one wrong decision.
What does a tracking setup that survives look like?
It has one clearly named conversion event per meaningful business action, sent from both the browser and the server with deduplication in place, respecting consent, and reconciled against the CRM on a fixed schedule. It is documented well enough that the next person can understand it, and it is tested after every website change.
None of that is exotic, and all of it is boring, which is precisely why it decays. Tracking is the only part of a digital marketing programme that silently stops working while every dashboard continues to look healthy. It is also the part that decides whether your Google Ads and Meta ads are learning from real outcomes or from noise, which is why we treat it as delivery work rather than as a technical afterthought inside LeadLock and every lead generation build.
If your reported conversions and your actual enquiries have stopped resembling each other, that gap is measurable and it is fixable. Tell us what you are seeing and we will show you where the numbers are leaking before you change a single campaign.