referrals git:20261009.7e8a78bA

Immutable. This exact content never changes and is served at /api/v1/blob/c2edbb33fb394de1.

---
name: referrals
description: Use when designing or fixing a customer referral program for an online store — the headline offer, who gets what reward, how big the reward is, cash vs gift, and the pre-filled share message — e.g. "design our refer-a-friend program", "should the reward be cash", "how much should we give per referral", "write the referral share message", "referrals aren't converting".
---

# Referrals

Design a referral program that gets more customers to refer and keeps referred customers profitable. The lever is framing: people refer because they think the friend will like it, and a reward aimed at the referrer's self-interest works against that.

## When to use

- The referral offer and headline; reward size and type
- The pre-filled share message
- A program with many referrals but low-value customers

## When not to use

- Reviews, ratings, or complaints → the `reviews` skill
- Loyalty rewards for repeat purchase → the `post-purchase` skill

## Inputs

Collect, or state your assumption:

- Current headline, rewards for referrer and friend, reward type, share message
- First-order margin (to size the reward)
- Goal: more referrals, more conversions, or better-quality customers

## Procedure

1. **Make the friend's benefit the headline.** "Give a friend $20 off", not "Earn $20 per friend". Keep a referrer reward, stated second and smaller. Altruistic framing made people up to 60% more likely to refer at least once; total referrals rose 86%.
2. **Pre-fill the share message with one signal.** Either "I already bought this" (+11% friend conversion) or "you can earn a reward for referring" (+53% onward referrals). Pick by goal. Never both. Separately from the persuasion signal, if the sender earns a reward the message must say so — a brief "(I get a thank-you if you use it)" — because US/UK advertising rules treat an incentivized recommendation without disclosure as deceptive; the studies predate this expectation.
3. **Reward in kind, not cash.** A free gift, upgrade, or other non-monetary benefit over cash or a cash discount; store credit as the fallback. †
4. **Cap the reward size.** Higher rewards raise volume but pull in poor customers: when an online bank raised its reward from €20 to €50, profitability of referred customers fell 48%. Sizing against first-order margin is a conservative practitioner policy, not the study's finding — with strong repeat value a larger reward can still pay; if you use lifetime value, use *incremental* profitability of referred customers. †
5. **Pay out after the friend's return window closes.** (Hygiene, not a study finding.)

## Output

**Program spec**

| Element | Value | Why |
|---|---|---|
| Headline | "Give a friend $20 off their first order" | Altruistic framing: +86% referrals |
| Friend reward | $20 off / free gift | Friend's benefit leads |
| Referrer reward | Free [upgrade/gift], shown second | Non-monetary; secondary |
| Reward size | Below the margin on a first order | High rewards → −48% profitability |
| Payout timing | After return window | Hygiene |

**Share message** (one signal):

> "I bought [product] and think you'd like it — here's $20 off your first order: [link] (I get a small thank-you if you use it)"

or, if the goal is chain referrals:

> "Here's $20 off [product] — and once you've tried it you can give friends $20 off too (I get a small thank-you for sharing this): [link]"

## Common mistakes

- "Earn $50 cash per friend" as the headline — lead with the friend's benefit.
- A share message with both "I bought this" and "you can earn too" — one or the other.
- Raising the reward for volume — referred-customer profitability drops.
- Cash payouts — gift, upgrade, or at minimum store credit.

## Quick reference

| Element | Do |
|---|---|
| Headline | Friend's benefit first, referrer reward second |
| Share message | One of: "I bought this" / "you can earn too" |
| Reward type | Gift or upgrade > credit > cash |
| Reward size | Modest; cap against first-order margin |
| Payout | After return window |

† No primary citation on file for this rule — a practitioner tip, not a measured finding.

Evidence and effect sizes: see [reference.md](reference.md).