# Referral Program

> Design a referral program that drives real word-of-mouth growth. Use when asked to build a referral or refer-a-friend program, create an incentive/reward structure, or turn happy users into a growth channel. Produces the incentive design (who gets what, when), the mechanics and trigger moment, fraud guardrails, and the unit-economics check — a program that pays back, not one that just burns budget.

- Skill: `mohitagw15856/referral-program` (Agent Skill)
- Install (CLI): `npx skillmds add mohitagw15856/referral-program`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mohitagw15856/referral-program/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Marketing & Growth
- Author: Mohit Aggarwal (https://skillmd.com/u/mohitagw15856)
- Updated: 2026-09-07
- Page: https://skillmd.com/skills/mohitagw15856/referral-program

---


# Referral Program Skill

A referral program turns happy customers into a growth channel — but most fail because the incentive is wrong,
the ask comes at the wrong moment, or the economics don't work. This skill designs one that does: the right
reward for both sides, the trigger at peak satisfaction, fraud guardrails, and a payback check so it's growth,
not a giveaway.

## Required Inputs

Ask for these only if they aren't already provided:

- **The product & economics** — what you sell, price/margin, and roughly your CAC and LTV (so rewards can be sized).
- **The "aha" / happy moment** — when users feel the value most (the right time to ask).
- **Audience motivation** — would they refer for cash, credit, status, or to help a friend? B2C vs B2B differs a lot.
- **Constraints** — budget per referral, legal/region limits, what's technically feasible.

## Output Format

### Referral program: [product]

**1. Incentive design** — who gets what, and when it pays out:

| Side | Reward | Triggers when | Why this reward |
|---|---|---|---|
| Referrer | | (e.g. friend's first purchase) | |
| Referred friend | | (e.g. on signup) | |

Double-sided usually beats one-sided. Reward the *outcome* you want (paid conversion), not just a click/signup.

**2. The ask moment & mechanics** — *when* to prompt (right after the aha moment / a great experience), where (in-product, email, post-purchase), and the share flow (unique link/code, how it's tracked, how rewards are granted). Keep it one or two clicks.

**3. The message** — a short, shareable framing the referrer would actually send (helping a friend, not spamming for a kickback).

**4. Fraud & abuse guardrails** — self-referral, fake accounts, reward farming; the checks (reward on real conversion, limits, verification).

**5. Unit-economics check** — total reward cost per successful referral vs. CAC and LTV. The program must acquire customers **below** your other channels' CAC (or clearly cheaper than paid) to be worth running. State the breakeven.

**6. Measure & iterate** — participation rate, referrals per advocate, conversion of referred users, and referral CAC vs. payback. What to tune.

## Quality Checks

- [ ] Incentive is sized against real CAC/LTV and rewards the outcome (conversion), not just a click
- [ ] The ask is triggered at a genuine high-satisfaction moment, with a low-friction share flow
- [ ] Double-sided vs. one-sided is a deliberate choice with a rationale
- [ ] Fraud/abuse guardrails are specified
- [ ] A unit-economics / breakeven check shows the program pays back
- [ ] Success metrics (participation, referral CAC, referred-user conversion) are defined

## Anti-Patterns

- [ ] Do not set a reward without checking it against CAC/LTV — that's just burning money
- [ ] Do not reward signups/clicks alone — reward the conversion you actually want
- [ ] Do not ask before the user has felt value — timing is half the program
- [ ] Do not ignore fraud — reward farming can quietly eat the whole budget
- [ ] Do not make sharing clunky — every extra step kills participation

## Based On

Referral/viral-growth practice (double-sided incentives, trigger-at-aha, referral CAC vs. LTV, fraud guardrails).