How to build a referral reward system from scratch in Africa

A complete guide to building referral reward programmes for African markets — covering attribution logic, trigger design, the right reward value per market, fraud prevention, multi-market delivery, and the Africa-specific decisions that determine whether the programme actually works at scale.

Referral programmes are among the highest-ROI customer acquisition mechanics available — a referred customer costs a fraction of a paid acquisition, has higher lifetime value, and churns at lower rates. African markets are structurally well-suited to referral mechanics: strong social networks, high trust in peer recommendations, and mobile-first communication channels that make referral sharing frictionless.

Yet most African businesses either don't run referral programmes or run them badly — with unclear attribution, delayed rewards, or fraud levels that eat the economics. This guide covers the full build: the mechanics, the API integration, the fraud considerations, and the market-specific decisions that separate referral programmes that grow from ones that flatline.

Step 1 — Define what constitutes a valid referral

The most consequential decision in a referral programme is what event validates the referral and triggers the reward. Get this wrong and you'll either over-reward (paying for referrals that don't produce real customers) or under-reward (missing legitimate referrals that should count).

In African markets, the three most common qualifying events are:

• Account activation with first transaction

  The referred person creates an account AND completes their first qualifying action — first purchase, first deposit, first payment. This is the highest-quality referral definition because it filters out sign-ups who never activate.

• First purchase above threshold

  The referred person completes their first purchase above a defined value — not just a registration. For e-commerce, fintech, and FMCG this is the right gate. Below the threshold, the acquisition cost exceeds the customer value.

• Verified account creation

  The referred person creates a verified account — phone number confirmed, identity checked at the level your programme requires. Appropriate for banking and financial services where even an unactivated account has compliance value.

Step 2 — Attribution: how to track who referred whom

Attribution is the technical backbone of a referral programme. You need to know, with certainty, that Customer B was referred by Customer A — so that Customer A's reward is correctly triggered when B qualifies.

In African markets, three attribution methods are practical:

Unique referral codes

Each existing customer receives a unique referral code — alphanumeric, 6–8 characters, generated at enrolment. They share this code with their network. New customers enter the code during registration. The code maps to the referrer's ID in your database.

Advantages: works across all channels including offline (a code on a flyer, spoken aloud, sent in a WhatsApp message). Simple for the referrer to share. No technical dependency on links or deep linking. This is the right approach for mass-market African programmes where referral happens through WhatsApp messages, verbal word-of-mouth, and informal networks.

Referral links

Generate a unique URL per referrer. When a new user clicks the link and signs up, attribution is handled automatically. Works well for smartphone-dominant, app-centric programmes.

Limitation in African markets: link sharing is less reliable when recipients share via WhatsApp voice messages, USSD, or verbal referral — which is common in informal trade and community networks. Always offer codes as an alternative even if you also support links.

Phone number attribution

During registration, ask the new customer "were you referred by someone? If yes, enter their phone number." The referrer's phone number becomes the attribution key. Slightly more friction than a code but works in markets where referrers don't have a code to hand but the new customer knows who told them about the product.

// 1. New customer registers with referral code
POST /api/register
{
  "phone": "+2348099887766",
  "referral_code": "QFT-A8821"   // maps to referrer ID
}

// 2. Store pending referral — not yet qualified
INSERT INTO referrals (
  referrer_id, referred_phone,
  status, created_at
) VALUES (
  'usr_88221', '+2348099887766',
  'pending', NOW()
)

// 3. Qualifying event fires (first purchase ≥ ₦2,000)
// Webhook from payment gateway confirms transaction

// 4. Look up pending referral for this phone number
SELECT referrer_id FROM referrals
WHERE referred_phone = '+2348099887766'
AND status = 'pending'

// 5. Issue reward to referrer — not the referred
POST https://api.qifts.com/v1/rewards/issue
{
  "recipient": { "phone": referrer_phone },
  "reward": { "value": 3000, "currency": "NGN",
               "category": "grocery" },
  "delivery": { "channel": "whatsapp" },
  "programme": { "reference": referral_id }  // idempotency
}

// 6. Mark referral as completed
UPDATE referrals SET status = 'completed' WHERE ...

Step 3 — Reward value calibration per market

Referral reward values need to be set at a level that motivates sharing without creating perverse incentives that attract fraud. The effective range varies by market and by the value of the referred customer:

The referred customer reward (if you choose to offer one) should be slightly lower than the referrer reward — you want the referrer to be the primary motivated party. A first-purchase discount or a welcome reward card for the new customer works well as a complementary mechanic without cannibalising the referrer's motivation.

Step 4 — Fraud prevention

Referral programmes in African markets attract organised fraud at meaningful scale. The typical fraud pattern: a single operator creates multiple SIM cards, registers multiple accounts using those numbers, completes minimum qualifying transactions between accounts, and claims multiple referral rewards. At ₦3,000 per referral, a fraud operation processing 100 fake referrals per day earns ₦300,000 before you notice.

The fraud prevention layer needs multiple signals working together: