# Referral Loop Audit

> Paste your refer-a-friend or affiliate program - the reward, where and when you ask, how people share, what the referred friend sees, the terms - plus price, LTV, gross margin, CAC and the program's numbers. The browser finds the four stages of the referral loop, holds the reward against the maximum the economics allow and flags the program's mistakes for free; a paid audit says what to fix, and a paid design writes the program.

URL: https://referral-loop-audit.skillsafe.ai/
API tutorial: https://referral-loop-audit.skillsafe.ai/api.html
Model: gpt-terra (metered per run; the free scan needs no account)

## Who it is for

Founders, growth leads and marketers whose customers like the product but rarely refer anyone, whose affiliate program signs up partners who never send a sale, or who are about to launch a program and want the reward sized before they promise it.

## What the free scan computes (in the browser, nothing uploaded)

- The program type (customer referral, affiliate or both), whether it runs today or is planned, and B2B or B2C.
- The referral loop, stage by stage, with the line that shows each: the trigger moment (after value or a milestone, at signup, or buried in settings, a footer or a menu), the share action (in-product, personal link, email invite, social, code; pre-written or manual), converting the referred friend (a landing that names who invited them, or the generic homepage), and the reward (who gets what, single- or double-sided or tiered, paid at signup or after a purchase, its type).
- The economics: the maximum reward, LTV x gross margin - CAC, and what one referred customer costs in rewards (referrer plus friend, free months and percentages priced from the stated price, lifetime commissions from LTV).
- The numbers: referral rate, active affiliates, referred customers per referrer, invite conversion, referred signups who pay, the K-factor, the share of new customers from referrals and the cost per referred customer.
- Affiliate terms: commission, recurring or lifetime, months, cookie window, buyout clause. Fraud protection, launch and reminder emails, terms.
- Flags RF-01..RF-20: reward paid at signup; single-sided; reward above the maximum; reward that cannot be sized; buried or missing ask; ask at signup; manual sharing; no fraud protection (high for cash or credit); uncapped cash rewards; a percentage reward on a product under about $100; a generic landing; referral rate under 10%; no program numbers; no emails; a lifetime commission with no buyout; a cookie under 30 days for B2B; a commission above the typical range; no terms; a product that spreads by use with no product-embedded mechanism; no reward at all.

## The paid lanes (one JSON input with a `task` field)

- `audit`: the program type and stage, the viral fit (natural, limited, unclear, with product-embedded mechanisms), the four loop stages (strong, weak, missing, each with a finding and a fix), the reward against the economics (maximum, cost, within / over / unknown), the numbers read against the skill's benchmarks, fraud protection with the gaps left open, issues as finding, impact and fix with severity, 3 to 6 fixes and a verdict (ship, fix first, redesign).
- `design`: the program type, 1 to 3 product-embedded mechanisms when the product spreads by use, 1 to 3 trigger moments (never at signup), the share mechanism and a pre-written message, the reward (structure, referrer, friend, framing, paid after the friend buys, cap, within the maximum), the landing for the referred friend, 3 to 5 program emails starting with the launch, 3 to 6 fraud rules, affiliate terms when there are affiliates, metrics (always referral rate), 2 to 4 experiments and a launch checklist. The audit hands its result to the design with one button.

Every reply is checked against the input: quotes word for word, no invented count, rate, price or window beyond the skill's constants, no blended lanes, and the lane-specific rules above.

## Sources

Derived from the agent skill @coreyhaines31/referrals (https://skillsafe.ai/skill/@coreyhaines31/referrals), by Corey Haines, in coreyhaines31/marketingskills (MIT): https://github.com/coreyhaines31/marketingskills/tree/main/skills/referrals - its SKILL.md and references (viral-mechanisms.md, program-examples.md, affiliate-programs.md).

Referral Loop Audit is not affiliated with or endorsed by Corey Haines, Rewardful, PayPal or any other tool named in the examples.

Part of SkillSafe: https://skillsafe.ai/