Referral Program
A referral program is the formal, repeatable mechanism a company uses to reward customers for inviting others, typically with discounts, credits, or perks. It operationalizes referral marketing into a trackable system.
Key takeaways
- Four decisions define a program: eligibility, trigger event, reward, and attribution window.
- Paying on signup maximises volume and fraud; paying on qualification maximises quality.
- Account credit costs less than cash and keeps the reward inside the product.
- The reward must stay below a referred customer's contribution margin to be sustainable.
- Self-referral, duplicate email and shared-payment checks belong in the build, not later.
In depth
A referral program is four rules plus a ledger. You decide who may refer, what event unlocks a reward, what the reward is, and how long a referred person stays attributed to their referrer. Everything else follows from those. Mechanically a referral moves through states: a link is issued, someone clicks it, an account is created and linked to the referrer, a qualifying event fires, and only then does the payout become owed. Each state needs a record you can audit.
The reward's trigger point sets the whole trade-off. Pay on signup and participation is high, gratification immediate, and the program attracts people who create accounts for the payout. Move the trigger to a qualifying event and quality rises while participation falls, because the referrer now waits and may never be paid. Reward type matters almost as much: account credit costs you less than cash and keeps the money inside the product, while status and early access often outperform both where the product carries prestige.
Building one means issuing a unique link per referrer, storing the referrer identifier on the referred account at creation, defining an attribution window, and adding checks for self-referrals, duplicate emails and shared payment details. Written terms should cap total payouts and state when a reward is revoked. A scorecard funnel gives an unusually precise trigger: make the qualifying event a completed quiz that scores into a chosen tier, and the reward is tied to lead quality rather than to a proxy for it.
A program formalises behaviour that must already exist. Where nobody recommends the product unprompted, adding a reward buys transactions rather than advocacy, and the referred accounts behave like paid acquisition with extra administration. The economics break whenever the reward exceeds the contribution margin of a referred customer over a realistic horizon, which is easy to miss when rewards are credits rather than cash. Programs also create standing liabilities, dispute handling and a rulebook someone has to maintain.
Example in practice
How to measure it
The core figure is cost per qualified referred customer: total rewards paid plus program running costs, divided by referred customers who reached the qualifying event. Compare it directly against your blended paid cost per acquisition. Track participation separately as the share of eligible customers who issued at least one link, because a cheap program that nobody uses looks efficient and contributes nothing.
Watch two integrity signals alongside the economics. The first is the distribution of referrals per referrer: a long tail is healthy, a handful of accounts producing most volume usually means gaming. The second is the gap between referred signups and referred qualifying events; if that gap widens, people are creating accounts to collect rather than recommending, and the trigger needs to move later.
Common mistakes
The common failure is launching without written terms and then improvising when the first dispute arrives. Someone refers thirty accounts from one address, a reward is withheld, and the argument happens in public. Publish eligibility, the qualifying event, the payout schedule, the cap and the revocation conditions before the first link goes out, and route every exception through one person rather than through whoever answers support.
The second is measuring the program on referrals generated. That number rewards the loudest incentive rather than the best one, and it rises fastest when the trigger is set at signup. Judge the program on referred customers who reach the qualifying event and are still active a quarter later, then divide total rewards paid by that count. If the result approaches your paid cost per acquisition, the program is not saving anything.
Frequently asked questions
What is the difference between one-sided and two-sided referral rewards?
One-sided rewards pay only the person making the referral, while two-sided rewards give incentives to both the referrer and the new customer. Two-sided rewards typically increase conversion because they lower the recipient's barrier to act.
Should referral rewards trigger on signup or on activation?
Trigger on activation or a qualifying action whenever possible. Rewarding raw signups inflates volume with users who never engage, while rewarding qualified actions protects your unit economics.
Should the reward be one-sided or two-sided?
Two-sided in most cases, because the referred person needs a reason to act on the recommendation and a discount gives the referrer a gift to offer rather than a favour to ask. One-sided rewards make sense when the product is free to start, since the recipient already has no barrier and the incentive would only add noise to their decision.
How long should the attribution window be?
Long enough to cover your normal consideration period and no longer. If most referred prospects decide within three weeks, a thirty-day window captures nearly all of them while a ninety-day one mostly captures people who would have arrived anyway. Longer windows also increase disputes, since a referrer can claim a customer who reached you through another channel entirely.
How do I prevent referral fraud?
Set the reward trigger past the point where a fake account is cheap to create, then add mechanical checks: matching email domains, shared payment instruments, identical device fingerprints, and referral counts far above the distribution. Cap payouts per referrer per period. Most abuse is deterred by a delayed trigger alone, because gaming stops paying when the reward requires real usage.
What should the qualifying event be?
The earliest point at which you are confident the referred person is real and a plausible customer. For self-serve products that is often a first meaningful action rather than a purchase; for sales-led ones it is usually a qualified opportunity. Setting it at account creation is fast and cheap to game; setting it at first payment is safe but slow enough to kill participation.
How large should the reward be?
Below the contribution margin a referred customer generates over a horizon you are willing to wait, and above the effort the referrer spends. Those two bounds often leave a wide range, so pick from the low end and raise it only if participation is genuinely the constraint. Raising rewards is easy and almost impossible to reverse without complaints.
Can a referral program hurt existing word of mouth?
It can, when the reward reframes a sincere recommendation as a paid one. People who were recommending freely may become self-conscious, and recipients discount an endorsement that came with a payout attached. The risk is highest for products with strong existing advocacy. Measure unprompted referrals before launch so you can tell whether the program added volume or simply relabelled it.