CPA looks cheaper than revenue share until you model the payback
The qualification bar is not an accounting detail. It is a specification handed to a counterparty who is paid to satisfy it, and it will be satisfied exactly and no further.
CPA is the easiest partner deal to sign and the hardest to evaluate. The cost is known on day one, the revenue arrives over a year, and the qualification criteria that trigger payment are the exact thing a partner optimises against. A broker can run a CPA programme for two years and never know which partners funded the business and which were funded by it.
Three structures, three different exposures
Every introducer deal is a choice about who carries the risk that a client is worth less than expected.
- Revenue share. The partner is paid a percentage of what the client generates. Risk sits with the partner, cost scales with value, and the broker cannot lose money on a client. Partners dislike it because payment is slow and uncertain.
- CPA. A fixed sum per qualifying client. Risk sits entirely with the broker. Cash goes out before any revenue arrives, and the entire question is whether the qualification bar predicts value.
- Hybrid. A smaller CPA plus a reduced revenue share, usually with a clawback window on the CPA if the client fails some condition. Splits the risk, and introduces the only mechanism that actually disciplines quality.
payback = CPA / ( ARPU_monthly × survival_rate )The denominator falls every month, so payback is not CPA divided by a constant. A cohort with a steep early survival curve can have an infinite payback period even though its first month ARPU looks healthy, and nothing in a monthly partner report will show that.
The bar is the product
A CPA deal pays on a defined event: a deposit above some amount, plus some traded volume, within some window. That definition is not an accounting detail. It is a specification handed to a counterparty who is paid to satisfy it, and it will be satisfied exactly and no further.
- 01Bar set at $250 and 1 lotwithin 30 days
- 02Partner acquires to that barnot beyond it
- 03Deposits cluster at $250and at exactly 1 lot
- 04CPA triggers on day 29paid in full
- 05Activity stops on day 31nothing further required
The diagnostic is distributional and it is unmistakable. Plot first deposit amounts for a CPA partner and for an organic cohort on the same axis. The organic distribution is smooth. The CPA distribution has a spike at the bar. The size of that spike relative to the tail above it is the clearest available measure of how much of the partner’s book is engineered.
Adverse selection in a CPA programme is not a partner problem. It is the contract working as written.
Blended partner reporting hides all of it
Partner performance is almost always reported as cost against revenue in the same period. That comparison puts this month’s CPA payments next to revenue generated mostly by clients acquired six months ago, which is not a ratio of anything. While a programme is growing it flatters. When it stops growing it collapses, and the collapse looks like a sudden deterioration in partner quality rather than the arithmetic catching up.
The only honest unit is the acquisition cohort: clients introduced by one partner in one month, tracked forward on their own. Everything useful about a partner is visible in the shape of that curve and invisible in any period total.
Both partners cost $42,000 in the month and both delivered 120 qualifying clients at $350. The difference appears in month four and is decided by month nine.
| Partner C, CPA | Partner A, rev share | |
|---|---|---|
| Qualifying clients | 120 | 120 |
| Cost in month one | $42,000 | $4,900 |
| Still funded at month 3 | 38% | 71% |
| Still funded at month 9 | 9% | 44% |
| Revenue per client, 12 months | $186 | $742 |
| Cohort revenue, 12 months | $22,320 | $89,040 |
| Total partner cost, 12 months | $42,000 | $26,700 |
| Cohort contribution | −$19,680 | +$62,340 |
Illustrative figures, chosen to show the mechanism. Your own numbers will differ.
What actually disciplines a CPA deal
Move the bar to where value is
If the qualification event does not predict a twelve month client, change the event. A bar that requires activity in a second month costs a partner more to satisfy and costs you far less to honour, because a client trading in month two is a genuinely different animal from one trading in week four.
Clawback windows that are actually reconciled
Most CPA contracts contain a clawback clause. Few brokers run the reconciliation that triggers it, because it requires matching a payment made months ago to a condition evaluated now, across a partner ledger and a client ledger that were never joined. An unenforced clawback clause is a clause the partner has correctly priced at zero.
Price the deal from the cohort, then renegotiate on it
- Sustainable CPA is a function of the cohort curve, not a market rate. It is the contribution you are prepared to pay away over the payback window, discounted for the survival you actually observe from that partner.
- A partner whose cohorts underperform does not need to be terminated. It needs to be moved to hybrid, which shifts precisely the risk that its acquisition method creates.
- A partner whose cohorts outperform is being underpaid and is being courted by someone else. The same analysis identifies both.
All of this needs one join that most brokers do not have standing: the partner attribution on a client, held permanently, against that client’s full revenue history across trading, swap and commission, grouped by the month they were introduced rather than the month the money arrived. That is the work.