Your customers like you. Why don't they refer anyone?
Paste your referral or affiliate program as it runs today or as you plan it - the reward, where and when you ask, how people share, what the referred friend sees, the terms - and a few lines on your price, LTV, gross margin and CAC and the program's numbers. Your browser finds the four stages of the referral loop, holds the reward against the most your economics allow (LTV x margin - CAC), computes the referral rate and the K-factor, and flags what leaks: a reward paid at signup, an ask buried in settings, a code people must copy by hand, a friend who lands on the homepage, no fraud rules, a lifetime commission. Free, before you sign in, and nothing is uploaded. Then the audit says what to fix, and the design writes the program: when to ask, the share message, the reward, the landing, the emails and the fraud rules.
The products and numbers in the examples are invented. Two are audits (the coffee one also carries its design, one click after the audit) and one is a design; each comes with a saved model reply, so you can see the whole page without signing in or spending a credit.
What this does, and what it does not
A referral program is a loop: ask at a moment when the customer has just felt the value, make sharing one click with a message ready, land the friend on a page that says who invited them and what they get, and pay the referrer once the friend buys. A leak at any stage starves the next one, and a reward bigger than LTV x gross margin - CAC loses money on every referral it wins. This page finds each stage in your program with the line that shows it, prices the reward per referred customer against that maximum, computes your referral rate and K-factor, reads the affiliate terms and the fraud protection, and raises the mistakes the skill warns about.
The paid lanes see your program (the middle is cut past 6,000 characters, and the page says so), your context (past 4,000) and the page's scan. Your input stays in this browser unless you run a lane. Every reply is held to your input: quotes are checked word for word, and no count, rate or price may be invented beyond the skill's own ranges - the design writes [placeholders] for what only you know. Derived from the agent skill @coreyhaines31/referrals by Corey Haines (coreyhaines31/marketingskills, MIT): its viral potential spectrum and seven viral mechanisms, the referral loop, incentive types and sizing, value presentation, the K-factor and benchmarks, affiliate commission, cookie and buyout mechanics, fraud prevention, the launch checklist, the email sequences and the tests to run. Sources are listed in llms.txt.