The campaign reports conversions. The dashboard is green. Nobody has put money in an account for three weeks. This is the most common way a budget disappears in restricted and regulated verticals, and it is almost never the creative.
A campaign has been running for a month. The reporting shows conversions climbing. Cost per conversion looks reasonable, sometimes impressive. Somebody sends a screenshot to the board.
Then the finance team asks a simple question: how many of those people funded an account? And the answer, uncomfortably, is close to none.
At that point the usual suspects get blamed. The creative was wrong. The audience was wrong. The offer was wrong. Sometimes they were. But in restricted and regulated verticals — gambling, crypto, betting, trading, forex, lending — there is a much more common explanation, and it has nothing to do with any of that.
The platform was never told where the money happens. So it optimised for something else, perfectly, for a month.
Every paid campaign optimises toward an event. You tell the platform "this action is a success", it goes and finds more people likely to perform that action. Straightforward, and it works remarkably well — as long as the event is in the right place.
In restricted verticals the traffic frequently passes through an intermediate page before it reaches the operator's own site. There are reasons for that, and we will come to them. The problem is where the conversion event is measured.
Toggle between the two setups below. Same campaign, same budget, same creative. Only the position of one line of tracking changes.
In the first setup the platform records a success the moment somebody clicks through. That person may never arrive on the operator's site. They may arrive and leave in four seconds. They may register and never deposit. None of that reaches the platform, because the platform already got what it was told to count.
In the second setup nothing is recorded until the person completes the action that actually matters — a registration on the operator's site, a qualified lead the sales desk can work with, a first deposit. Fewer events, later events, and every one of them true.
This is worth saying plainly, because it is rarely malice and usually structural.
An agency or a media buyer generally cannot place tracking on the operator's own site. Sometimes because access is not given, which is entirely reasonable — a business does not hand its analytics to a supplier it met last month. Sometimes because the operator's technical team has a queue three weeks long. Sometimes because the vertical makes anyone nervous about what gets connected to what.
So the buyer measures what they can reach. The intermediate page is theirs, so the event goes there. It is the path of least resistance, and it produces a report that looks fine.
Nobody sets out to mislead the algorithm. They just measure the only thing they were allowed to touch.
And there is a second reason, less comfortable. A campaign optimising for clicks produces a lot of cheap conversions. A campaign optimising for deposits produces fewer, more expensive, more honest ones. If the relationship is judged on a cost-per-conversion figure in a monthly report, the first version wins every review — right up until somebody asks about revenue.
Meta's delivery works in two stages across Facebook, Instagram and Threads. Andromeda narrows an enormous pool of eligible adverts down to what is worth considering for a given person. GEM then ranks what survived and decides who actually sees it. Both stages improve using the outcomes you report back.
Report the wrong outcome and both stages get better at the wrong thing. Precisely. Relentlessly. The system is not broken — it is doing exactly what it was asked.
The practical effect looks like this:
Three months of this and the audience the account has learned to reach is genuinely the wrong audience. Fixing the tracking at that point does not reset it instantly — there is relearning to do, and it costs time as well as money.
Here is the part that gets left out of articles like this one, because it complicates a clean story.
Optimising toward a deeper event is correct. It is also harder to feed. Delivery systems need a reasonable volume of events to leave the learning phase and stabilise. Deposits are rarer than clicks by definition. Qualify aggressively — long forms, pre-qualification questions, strict criteria — and the quality of each lead rises while the number of events the system receives falls.
Push that too far and the campaign never stabilises at all. You end up with excellent leads and an algorithm that is permanently guessing.
Getting that balance right is not a setting. It depends on your volume, your margin, your market and how long your sales cycle is — and it moves as the account matures. An agency that tells you the answer before looking at your numbers is guessing.
The reasonable objection to all of this: you are not going to give a new supplier access to your customer data, and you should not have to.
You do not have to. Modern conversion tracking is designed for exactly this situation — events can be sent from your own server, under your control, containing only what you decide to send. Your analytics stay yours. What reaches the platform is a signal that something valuable happened, not a copy of your database.
Setting that up properly in a restricted vertical takes a technical conversation rather than a checkbox, and the details vary enough per operator that publishing a recipe would be useless. What matters is knowing it is possible — because the alternative most accounts are running is not a technical limitation, it is a shortcut nobody re-examined.
Fewer conversions in the report, more money in the account. That trade is the entire job.
Three questions worth asking whoever runs your campaigns this week:
If the answers are vague, that is the finding. It is also fixable, and usually faster than rebuilding creative that was never the problem.
Related
High-risk media buyingOpen → Forex marketing, measured on depositsOpen → CRO, tracking and attributionOpen → Case files with the actual numbersOpen →Less Bureaucracy.
More Work.
Specialists assigned to your account, decisions explained before they run, and a number at the end that means something. Start the conversation — it costs nothing and you will leave it knowing more than you arrived with.
Published August 2026 · More articles