DM9: Growth engineering: referrals, sharing and lifecycle messages
Referral programmes that resist abuse
🎯You'll be able to
Design a referral programme with fair rewards, clear rules and controls against self-referral and fraud, and work out what it costs.
🏠The everyday picture
A coaching class that offers a fee discount for each new student a student brings will soon see "new students" who are the same student under a different name, unless someone checks.
📘What it really is
-
Parts: a unique referral code or link per customer, a reward for the referrer, often a reward for the friend, and rules for when rewards are earned.
-
Reward timing: reward after the friend's first order is delivered and the return period has passed, not at sign-up.
- Abuse to expect:
- self-referral (one person, several accounts, phones or emails);
- coupon sites posting codes publicly;
- fake orders that are later cancelled or returned;
-
bots creating accounts.
-
Controls: one reward per device, payment method and delivery address; limits per referrer per month; delayed rewards; reviews of unusual patterns; clear terms. Use only the data you need for these checks, and say so in the privacy notice (DM4).
-
Cost: reward cost × referred customers, plus fraud losses, compared with CAC from other channels and the referred customers' value.
- Measure incrementality: some referred friends would have bought anyway; a holdout (no referral offer for a random group) shows the extra customers the programme brought (DM7, DM8).
📝A worked example
Input: an illustrative month: 4,000 buyers, 300 referral links used, 180 friends' first orders delivered, 30 later cancelled or returned; ₹100 to each side for delivered, kept orders.
Steps:
- Rewarded referrals: 180 − 30 = 150; cost: 150 × ₹200 = ₹30,000.
- Cost per referred customer: ₹200, against a search ads CPA of ₹125 (DM8 example).
- Check incrementality and the referred customers' repeat orders before judging.
Expected output: a note that referrals look more expensive per customer, so the decision depends on whether referred customers return more often and how many were extra. (Illustrative figures.)
🌍Where it's used
- Fintech apps run invite rewards with strong fraud checks.
- Food delivery and ride apps give credits to both sides.
- Software companies reward referrals with free months.
📝Case scenario
A grocery app gave ₹150 credit on sign-up through a referral link. Within a week, one address had created 40 accounts. The app moved rewards to after the first delivered order, limited rewards per address and payment method, and the abuse fell sharply.
🖱️Try it
Write the rules for a two-sided referral programme for a business you know: who gets what, when, and three limits. Success: each limit blocks a specific abuse.
⚠️Common mistakes
- Rewards at sign-up. Sign: accounts created and abandoned. Fix: reward after a delivered, kept first order.
- No limits. Sign: one person earning dozens of rewards. Fix: per-device, per-address and per-month limits.
- Ignoring cost per customer. Sign: "Referrals are free growth." Fix: compare with other channels and measure incrementality.
🛠️For your project
Your capstone's design notes can describe a MittiMart referral feature with its rules, limits and measure.
✅Quick check
When should a referral reward be paid?
Show the answer
After the friend's first order is delivered and kept, not at sign-up.
Name two controls against referral abuse.
Show the answer
For example limits per device, address or payment method, monthly caps, and delayed rewards.