How to set KPIs for CPA campaigns: FTD, approve rate and retention
A CPA deal is only as good as its KPI definition. This guide explains how to define billable events that both sides can validate — from FTD and qualified leads to approve rate, retention cohorts, holds and attribution windows.
What a KPI actually is in a CPA deal
In CPA marketing, a KPI is not a dashboard metric — it is the contract. It defines which user action triggers a payment, under what conditions, and who judges validity. Every other term in the deal — payout, caps, hold, attribution window — exists to serve that definition. When offer owners say "our KPI is FTD," they usually mean the headline; what the buyer hears is a dozen unanswered questions about qualifiers, validation and windows. The gap between the headline and the full definition is where disputes are born.
A complete KPI definition has five parts. The event: the user action that is billable. The qualifiers: minimums and conditions — deposit size, new-customer status, GEO, completed verification. The attribution window: how long after the click the event still counts. The source of truth: whose data decides validity. And the feedback loop: how and how fast validation results return to the buyer. Write all five down before traffic starts; any one left implicit will be negotiated later, with money already spent.
FTD: the workhorse KPI of iGaming and trading
FTD — first time deposit — is the standard billable event in casino, betting, forex and crypto because it is the first moment the user has cost the platform nothing and paid it something. A typical FTD definition: a first deposit of at least an agreed minimum, by a new verified customer, within an agreed window after the click, confirmed in the advertiser's payment system. Each part does work. The minimum filters out token test deposits. The new-customer condition prevents paying for existing users. The window bounds attribution. The payment-system confirmation is the source of truth.
The FTD is also where quality games begin, so define it against the abuse you have actually seen. Bonus hunters make minimum deposits to trigger payouts and never play — counter them with activity qualifiers (a wager requirement, a second deposit) if your economics need them, but know that every qualifier lowers the buyer's conversion rate and raises the payout they need. Multi-accounting passes naive new-customer checks — counter it with device and identity signals in your validation, and share rejections with reasons. The right FTD definition is the deepest one your validation can actually enforce quickly.
| KPI | What it pays for | Main risk | Best for |
|---|---|---|---|
| Registration (SOI) | Form submit | Bots, incent, zero intent | Volume testing only |
| Qualified lead (DOI+) | Verified registration | Weak downstream value | Dating, sweeps, lead-gen |
| FTD | First deposit ≥ minimum | Bonus hunters, multi-accounts | iGaming, trading, crypto |
| Sale / COD confirm | Paid or confirmed order | Refunds, non-delivery | E-commerce, nutra |
| Activation | First meaningful use | Long validation lag | Fintech, subscription apps |
Approve rate: the KPI that decides the payout
Approve rate — the share of delivered conversions that pass your validation — is not one KPI among many; it is the number the payout is calculated against. The arithmetic is unforgiving. If your payout is $50 per confirmed order and your approve rate is 40%, the buyer earns $20 per delivered conversion. Every point of approve rate you improve is margin for both sides; every point it falls is a negotiation you will have later.
The drivers of approve rate sit mostly on the offer side. Confirmation speed: in COD verticals, leads called within minutes confirm at multiples of leads called hours later. Delivery and payment coverage: orders fail where your courier or payment methods do not reach. Expectation match: funnels that over-promise produce conversions that users abandon. And only then traffic quality — which is real, but measurable: share validation results per source, and weak sources are filtered within days. What destroys the loop is slow or unexplained validation. If conversions sit unreviewed for two weeks, the buyer cannot optimize, and the approve rate you finally report describes sources that no longer run.
Retention: the KPI that separates growth from churn-buying
Validation tells you the conversion happened; cohorts tell you whether it mattered. A source that delivers first deposits with zero repeat deposits, registrations with zero week-one activity, or activations with zero retention is selling you expensive noise — and naive validation pays for it happily. This is why mature CPA relationships gate scaling on cohort metrics: repeat deposit rate in iGaming, registration-to-sale rate in dating, activation and transaction rates in fintech, refund-adjusted delivery in e-commerce.
Sharing cohort data per source feels like giving away leverage; in practice it is the opposite. It converts the buyer from a volume vendor into a quality partner — sources get optimized toward users who generate revenue, because the buyer can finally see which those are. The alternative equilibrium is worse for you: the buyer, blind to downstream value, optimizes for the cheapest conversions that pass validation, which are rarely the valuable ones. Define the gating metrics in the written terms — "volume scales when the previous cohort's repeat deposit rate holds at X" — and the incentives align without a single awkward conversation.
Holds, windows and the speed of truth
Two time parameters shape every CPA deal: the attribution window, and the hold. The attribution window defines how long after the click a conversion is billable — too short and you underpay for real influence; too long and you pay for users who would have converted anyway. The hold defines how long validation may take before payment — it exists because real validation (cleared deposits, confirmed deliveries, fraud checks) takes time. Both are commercial terms, and both belong in writing.
The principle that keeps both healthy: the speed of truth. Validation that takes weeks makes optimization impossible — sources cannot be judged, good ones are underpaid, bad ones over-deliver. Set the hold to the shortest window your operations can genuinely meet, then meet it every week. If your call center needs 48 hours to confirm orders, hold is 48 hours plus reconciliation, not "about two weeks." Buyers plan risk around the hold you commit to; an honest short hold earns better terms than an optimistic long one.
The five KPI mistakes that kill deals
First: the vague headline. "We pay per deposit" with no minimum, no window, no validation rule — discovered as a disagreement at the first reconciliation. Second: the unenforceable qualifier. A wager requirement your system cannot report per conversion is not a qualifier; it is an argument. Third: the silent rejection. Conversions declined without reasons, so the same "fraud" keeps arriving because nobody could filter it. Fourth: the retroactive change. Terms tightened mid-campaign and applied to conversions already delivered — the fastest way to lose every serious source you had. Fifth: the frozen hold. Validation that takes as long as it takes, because no one owns it.
Each mistake has the same shape: something left implicit becomes a conflict, and the conflict is settled against the relationship. The KPI and payout terms page describes how these terms are fixed in writing when you work with us, and the launch checklist turns them into twenty concrete preparation points.