Referral Program Design

Choose the incentive structure (double-sided, single-sided, tiered), the reward type, and the eligibility rules before launch.

Why

The incentive structure determines whether the program feels generous or extractive — double-sided ('you both get $10') converts 3-5x single-sided ('you get $10, they get nothing') because the inviter doesn't feel like they're profiting off a friend. The reward type must match the product: a product credit works for SaaS (it reduces churn risk on the inviter), cash works for consumer marketplaces, status works for communities. Eligibility rules (new-users-only, one-per-IP, geo-restrictions) prevent the fraud that kills program economics within a week of launch. Designing these upfront is far cheaper than redesigning after the first fraud wave or after the 'my friend felt used' complaints.

When

When organic k > 0.2 and you want to amplify it with structure. Design once, iterate the incentive size quarterly.

How

1. Pick the incentive structure: double-sided (both get reward) for consumer, single-sided for B2B where the inviter's reward is status
2. Pick the reward type: product credit (SaaS, reduces churn), cash (marketplace), feature unlock (freemium), status (community)
3. Size the reward to be < CAC but > the friction of inviting: if CAC is $20, a $10 double-sided reward costs $20 total and breaks even
4. Write eligibility rules: recipient must be new (never signed up), one reward per household/IP, geo-restrictions where tax/legal differs
5. Build the tracking: unique referral code per user, attribution window (30d cookie), reward fulfillment (instant credit vs delayed)
6. Add fraud signals to the spec: velocity (5+ signups/hour from one code), disposable-email rate, same-IP clusters
7. Launch to 10% of users first; measure k, fraud rate, and cost-per-acquired-user before full rollout
8. Iterate quarterly: if k is below target, raise the reward; if fraud is high, tighten eligibility; if cost/acq > CAC, lower the reward

Example

# SaaS note app — referral program spec
Structure: double-sided (inviter +10% off, recipient +10% off first 3 months)
Reward type: product credit (reduces inviter churn, no cash out the door)
Size: $6 value each side → $12 total, CAC is $18 → saves $6 per acquired user
Eligibility:
  - recipient: never signed up before (checked by email + device fingerprint)
  - one reward per household (IP + shipping address for any paid plan)
  - geo: US, CA, EU, UK, AU (tax handled); excluded elsewhere
Attribution: 30-day cookie on the referral link, last-touch wins
Fraud signals:
  - 5+ signups/hour from one code → auto-pause code, manual review
  - disposable email domains → block at signup
  - 3+ signups from same IP within 24h → flag, withhold reward

# Launch
10% rollout (Wed), measure 2 weeks:
  k: 0.22 → 0.31  ✓
  fraud: 1.2% of rewards  ✓ (under 3% threshold)
  cost/acq: $12 (vs $18 paid CAC)  ✓ → full rollout

Common Problems

⚠Single-sided reward where the inviter gets paid but the friend gets nothing — the inviter feels like a grifter and the friend feels used.
⚠Cash rewards in a SaaS — they attract money-motivated inviters who churn, and the accounting/tax overhead eats the margin.
⚠No eligibility rules, so one user with 50 email addresses collects 50 rewards in a weekend and the program is 'unprofitable' forever after.
⚠Launching at 100% and discovering the fraud rate is 15% only after paying out $20k in fraudulent rewards.

Best Practices

✓Use double-sided rewards ('you both get X') — they convert 3-5x single-sided and avoid the 'profiting off a friend' discomfort.
✓Match the reward type to the product: product credit for SaaS (reduces churn), cash for marketplaces, status for communities.
✓Write eligibility rules before launch and auto-pause codes that trip fraud signals; fraud prevention is 10x cheaper than fraud recovery.
✓Launch to 10% for 2 weeks and measure k, fraud rate, and cost/acq before full rollout; a bad program is expensive to unwind.