Build or buy

The parts of affiliate tracking that are harder than they look.

A referral link and a percentage is a weekend of work. Retry-safe intake, versioned commission rules, refund reversal and approved payouts are the parts that take a year and never stop needing attention.

01ReferralLink or code
02ConversionStable event
03CommissionRule evidence
Idempotent intakeVersioned rulesReversal semantics

Built for operating reality

Clear contracts at every boundary.

01

The first version always works

A code, a cookie and a percentage of the first invoice is genuinely simple. Nothing in that version is wrong until money has to move and somebody asks why a number is what it is.

02

The cost arrives with the second year

Renewals, partial refunds, chargebacks, plan changes, currency, duplicate webhooks and disputed attribution each arrive separately, and each one lands on whoever built it.

03

Buy the semantics, not the dashboard

The dashboard is the cheap part to build. The expensive part is deciding what a retried webhook means, and being able to prove that decision a year later.

The honest case for building

Build it when the programme is the product.

If commission logic is a differentiator your customers choose you for, own it. If your programme is one affiliate tier, one currency and no refunds, a table and a scheduled job will serve you for a long time. Neither case needs a vendor, and a vendor who tells you otherwise is selling.

The honest case for buying

A timeout is unknown, not permission to pay twice.

Payment providers retry. A conversion recorded twice becomes a commission paid twice, and the reconciliation is manual. Getting that right needs stable external identities, idempotency keys, a bounded conflict response and a ledger that balances. That is the work you are actually buying.

Questions, answered

Know where the boundaries are before you integrate.

How long does building affiliate tracking really take?

A working first version takes days. The durable version is bounded by the failure cases rather than the features: duplicate delivery, refunds after payout, rule changes applied to historical commissions, and a payout path that cannot double-pay. Budget for those or expect to discover them in production.

What breaks first in a home-built programme?

Usually refunds. A commission accrued on a payment that is later reversed has to be clawed back, and if the payout already ran the correction has to carry into the next cycle without losing the audit trail.

Can we start in-house and migrate later?

Yes, and it is a reasonable sequence. Keep stable external identities for customers and transactions from day one — migration is straightforward when those exist and painful when attribution was only ever a cookie.

Build the reliable path

Make every referral decision traceable from click to payout.

Request sandbox access