A new payment product in Almaty runs ads in two channels, each spending 200,000 ₸. Channel A brought 100 sign-ups and B only 50, so the marketer proposes moving the budget to A. Sabina from the product team opens the next step: 20 people from A and 30 from B paid for a first purchase. Cheap sign-ups do not pick the channel for you.
Take a debit product for paying for purchases. All the numbers in this lesson are illustrative, including the bonus and the product economics. Your user already has a payment app, Kaspi for example. They may sign up for a new one out of curiosity, but they will only pay with it once they understand why. For a loan, a deposit or an investment product, the result and the costs are defined differently; this calculation is about paying for purchases.
Name the action the person came for
A sign-up creates a record in the system. Identity verification is the document check without which the product will not let anyone pay. The first purchase payment shows that the person has started using the product. These events happen at different times, and each may be handled by a different employee.
In the report, a first payment is a completed purchase from a merchant within 30 days of sign-up, confirmed by the transaction record. Test and cancelled transactions do not count, and a repeated notification about one transaction does not create a new one. A person counts once, however many purchases they make.
Everyone in both groups has been observed for the full 30 days.
Cost per user with a first payment
| Metric | Channel A | Channel B |
|---|---|---|
| Spend | 200,000 ₸ | 200,000 ₸ |
| Registrations | 100 | 50 |
| Users with a first payment | 20 | 30 |
| Spend per such user | 10,000 ₸ | 6,667 ₸ |
Everyone completed 30 days. The first purchase is confirmed by the transaction record; tests and cancellations are excluded, each person counts once.
The "payment complete" screen is for the user; the analyst checks the event against the transaction record. If a refund arrives, the result is recalculated by a rule written down in advance. Confirmation comes from one source: the app screen and the payment ledger easily give different totals, and you do not want to be working out which is right in a meeting with the client.
Price each result on the same cost base
| Metric | Channel A | Channel B |
|---|---|---|
| Ad spend | 200,000 ₸ | 200,000 ₸ |
| New unique sign-ups | 100 | 50 |
| Cost per sign-up | 2,000 ₸ | 4,000 ₸ |
| Users with a first payment within 30 days | 20 | 30 |
| Share of sign-ups | 20% | 60% |
| Ad spend per user with a first payment | 10,000 ₸ | About 6,667 ₸ |
A's sign-ups cost half as much, but its first payments cost more. Still, these are two groups you observed once. Moving the whole budget to B would also need to know how much suitable audience is left in the channel, how people use the product afterwards and what serving them costs the product.
So far the table holds only advertising. If one partner was also paid a commission and a salaried team helped another, those costs go alongside. Full CAC is built from items agreed in advance. If you compare one channel's full CAC with another's bare ad budget, the choice is skewed from the start.
Find the obstacle between sign-up and payment
People can stop short of paying for many reasons: they did not understand the terms, could not upload a document, are waiting for a check, or simply do not plan to pay this way yet. An "inactive" status will not tell anyone what to fix.
Walk the path with Sabina on a test account and answer four questions. Is it clear what to do after signing up? Does what you filled in survive closing the app and coming back? Does an error message explain what to fix? Can you write to support without retelling the whole story?
Customer verification decisions belong to the company's specialists, and the marketer stays out of them. Form usability is one conversation, the content of the check another. FATF guidance treats a risk-based approach together with access to financial services. What matters for you is telling a technical difficulty apart from a verification outcome: fixing document upload is no reason to promise in an ad that everyone will pass.
A loan product would add one more decision, whether to grant the loan. Passing identity verification does not approve a loan. Before writing about statuses in an ad, ask the process owner what each one means.
First purchase and the habit of using the product
Someone may meet the bonus condition and forget the product. That is why, after the first payment, you check whether a purchase repeats within the same window.
Everyone who paid for the first time has had another 30 days from that payment. In A, 12 of 20 made another purchase; in B, 18 of 30. That is 60% of starters in both channels. Measured against all sign-ups the shares differ: 12 / 100 = 12% and 18 / 50 = 36%.
| What we count | Channel A | Channel B |
|---|---|---|
| Users with a repeat purchase | 12 | 18 |
| Share of users with a first payment | 60% | 60% |
| Share of all sign-ups | 12% | 36% |
| Ad spend per user with a repeat purchase | About 16,667 ₸ | About 11,111 ₸ |
The repeat window runs from the first payment. Someone who first paid on day 30 after sign-up needs another 30 days of observation, and until then their result is marked immature. The rules of retention analysis must be the same for both channels.
The same 60% can describe very different product lives. One person pays for lunch every day, another buys electronics once a month. Which repeat matters for your product is the team's call, and then they compare it with support requests and income.
The bonus is part of acquisition cost
In channel B, each of the 30 users with a first payment was credited and paid 3,000 ₸. Bonuses cost 90,000 ₸, or 290,000 ₸ together with advertising. Per user with a first payment that is 290,000 / 30 ≈ 9,667 ₸.
The same money over 18 users with a repeat gives about 16,111 ₸. Spend did not grow; you are dividing it by a different result. Write the bonus terms so the person knows in advance which action earns it, when it becomes available and where it can be spent. The Law On Advertising (as amended up to 17 May 2026) prohibits concealing material contract terms named in an advertisement (Art. 14, subpara. 5), and a licensed advertiser states the licence number and the body that issued it (Art. 6, para. 4).
In the example the bonus has already been paid. If it is only promised, future payouts are tracked next to the money already spent. An unpaid bonus makes a campaign cheap only on paper: the obligation stays after the ad report is closed.
Purchase volume and product income
Money a person pays by card in a shop passes through the product but does not become its income. A top-up is not company revenue either. What counts as the product's income and costs is decided by the finance team, and you take that definition from them.
Suppose that for group B over a closed period finance confirmed 100,000 ₸ of income and 40,000 ₸ of variable transaction and servicing costs. Advertising and bonuses are not in these figures. Contribution before them is 60,000 ₸.
After 200,000 ₸ of advertising and 90,000 ₸ of bonuses, the period result is 60,000 - 290,000 = minus 230,000 ₸, before other costs. Future transactions may change that, but they are forecast and later checked against actuals. The number of first payments will not close this gap in money.
If the product offers several services, income and shared costs are split by the company's rule. The same money must not fully pay back several campaigns at once.
Data for the report and customer data
The marketer needs the source, the agreed event, the time, the processing status and group totals. Document copies and the reasons behind decisions on each customer stay with the people responsible for them. They do not go into ad event parameters or their labels.
Before integration, the data owners agree on fields and recipients, and a developer runs test transactions, cancellations and repeated notifications. If a send was delayed or rejected, the report must show a data error. An empty event in the ad account may mean a failed transfer, so it is too early to file that customer as inactive.
Exercise: justify the next budget
The owner proposes doubling channel B because a first payment there costs about 6,667 ₸. Put together a short calculation for him with the bonus and repeat purchases, and list the data still missing for the decision.
With bonuses, channel B cost 290,000 ₸. That is about 9,667 ₸ per user with a first payment and about 16,111 ₸ per user with a repeat, and the period economics come out at minus 230,000 ₸. The 6,667 ₸ figure counts only advertising and only the first payment, and a doubling cannot rest on it.
A larger share of B's sign-ups reach a first and a repeat payment than A's, which is an argument for B, not yet for doubling its budget. Doubling the budget also needs future income from these people, the cost of serving them, how much audience is left in the channel and how much the company is willing to put in before the money comes back. Write in the report the date you will return to the decision and what actual data should be in by then.