mbfinotti/partnerships-skills已通过检查
SKILL DETAIL
referral-incentive-design
mbfinotti/partnerships-skills/referral-incentive-design
Design the reward structure of a customer refer-a-friend program - who gets rewarded (referrer, friend, or both), reward type (cash, account credit, discount, product value, status), how much to pay against LTV/CAC and gross margin, when the payout fires, and whether to build a tiered reward ladder. Use whenever the user mentions referral incentives, refer-a-friend rewards, referral bonus amounts, give-X-get-Y offers, double-sided referral rewards, or asks how big a referral reward should be, even if they never say incentive. Covers end-customer rewards only. Do NOT use for commission rates paid to professional affiliates - use mbfinotti/partnerships-skills@affiliate-commission-structure instead.
安装量 · 507查看来源
Installation
npx skills add https://github.com/mbfinotti/partnerships-skills --skill referral-incentive-design
技能文件
SKILL.md
最近同步 · 2026年10月23日
evals/evals.json›
{
"skill_name": "referral-incentive-design",
"evals": [
{
"id": 1,
"prompt": "I run Pinelope, a meal-planning app. The paid plan is $9/month, gross margin is 88%, average customer stays 11 months, blended CAC is $34. About 12% of new signups tell us a friend mentioned us. We want to launch refer-a-friend next month and I'm planning 'Give $10, Get $10' on both sides because that's what everyone does. What should the rewards be?",
"expected_output": "A reward structure spec that computes the gross-profit ceiling, shows the proposed $20 total is too large against it, leads with a friend-weighted or friend-only structure instead of the symmetric default, frames the friend reward as a free month rather than a dollar amount because the price is under $100, sets a paid trigger, and includes a sidedness test plan with an incrementality haircut.",
"files": [],
"expectations": [
"Computes a gross-profit ceiling as LTV times gross margin times an allowed acquisition percentage, showing the arithmetic (LTV around $99, gross profit around $87).",
"States whether the proposed $10 + $10 = $20 total sits inside or outside the ceiling it computed, rather than accepting the amount without testing it.",
"Recommends a friend-only or friend-weighted structure as the leading option instead of accepting the symmetric double-sided default the user proposed.",
"Justifies that sidedness lead with the field-experiment result that recipient-only rewards matched double-sided rewards costing about twice as much.",
"Names the Reward-Me / Reward-You / Reward-Both sidedness taxonomy (Ryu & Feick) explicitly.",
"Applies the Rule of 100: because the plan price is under $100, the friend's reward is framed as a percentage or a free month rather than as a dollar amount off.",
"Proposes a concrete friend-side framing such as \"first month free\" instead of \"$9 off\" or \"$10 off\".",
"Sets the payout trigger on a verified paid event, not on the referred friend's signup.",
"Includes a validation plan naming test arms across sidedness (both-sides / friend-only / referrer-only) run on a small share of the active base.",
"Applies an incrementality haircut of roughly half before comparing cost per referred customer against the $34 blended CAC.",
"Delivers the design as a structured reward-structure spec with explicit lines for sidedness, reward per side, payout trigger, tiers, caps, economics, and validation."
]
},
{
"id": 2,
"prompt": "Hollowmast is a B2C photo-printing subscription at $24/month, 72% gross margin, typical tenure 9 months. Our CFO will only approve a very small pilot budget, so the plan is to give referrers a $3 account credit per friend who subscribes, nothing to the friend, and raise it later if it works. Sanity-check the number for me.",
"expected_output": "A rejection of the $3 reward as sitting below the motivation floor, citing the non-monotonic response curve, with the pay-nothing-or-pay-meaningfully rule, a recognition-only alternative that fits the budget constraint, a ceiling calculation showing a real reward is affordable, and advice to fix the ask before touching the number.",
"files": [],
"expectations": [
"Rejects the $3 reward as sitting below a motivation floor, rather than treating it as a safe small starting point to iterate upward from.",
"States the non-monotonic result explicitly: introducing a small payment can produce worse performance than no payment at all.",
"Attributes that result to the \"Pay Enough or Don't Pay at All\" finding (Gneezy & Rustichini) or names the mechanism where a token payment converts an intrinsically motivated favor into a badly paid job.",
"Gives the binary rule: pay nothing and lean on recognition, or pay meaningfully, but never pay a little.",
"Offers a zero-cost or near-zero-cost alternative (recognition, status, early access, or an unpaid ask) as a legitimate option consistent with the CFO's budget constraint, instead of only advising more spend.",
"Recommends fixing the placement and timing of the ask before raising the reward size, noting placement moves participation more than reward size does.",
"Flags referrer-only as the least efficient sidedness arm, and that giving the friend nothing leaves out the side with the harder job.",
"Computes the gross-profit ceiling from $24/month, 9-month tenure and 72% margin (roughly $155) to show a meaningful reward is affordable and budget is not the binding constraint.",
"Proposes a specific, meaningfully sized reward amount inside that ceiling rather than leaving the size open.",
"Does not propose a staged test plan that launches at $3 and steps the reward up if it underperforms."
]
},
{
"id": 3,
"prompt": "Tidewrack is a $79/month B2B project tool, 83% gross margin, average tenure 19 months, paid CAC $410, blended CAC $290. The proposed structure is: friend gets one month free, referrer gets $60 in account credit. That's $139 against a $410 CAC, so it looks like a 3x win. Finance wants a one-page justification. Can you write the economics?",
"expected_output": "An economics section that separates nominal from real cost, computes the gross-profit ceiling, applies the incrementality haircut to produce a cost per incremental referred customer, benchmarks it against blended rather than paid CAC, states the pass threshold, and specifies a holdout as the verification method.",
"files": [],
"expectations": [
"Distinguishes nominal reward cost from real cost, stating that account credit costs the margin on redemption rather than its face value.",
"Computes a gross-profit ceiling from LTV (around $1,501) times the 83% gross margin times an allowed acquisition percentage, reports the resulting dollar ceiling, and says whether the structure sits inside it.",
"Applies an incrementality haircut of roughly 50%, on the basis that about half of rewarded referrals in the research would have converted anyway.",
"Reports a cost per incremental referred customer as a figure distinct from the cost per successful referral.",
"Benchmarks that incremental figure against the $290 blended CAC rather than only against the more flattering $410 paid CAC.",
"Explicitly corrects the user's framing that $139 against $410 is a 3x win, showing the raw comparison overstates the advantage.",
"States a pass threshold the structure must clear: cost per incremental referred customer below current blended CAC and inside the gross-profit ceiling.",
"Specifies a holdout test, withholding the program from a slice of customers or a geography, as the way to verify incrementality.",
"Prefers the holdout over the crude pause or on/off check, or names the on/off check only as the fallback when the base cannot be segmented cleanly.",
"States how the cost is booked: referrer rewards as customer-acquisition cost in sales and marketing, friend-side discounts as margin give-up, and unredeemed credit as a liability.",
"Notes that referred customers carry a measured LTV premium and lower churn, which is what justifies paying near the top of the range."
]
},
{
"id": 4,
"prompt": "Quillbarrow sells a $140/year consumer backup service with a 60-day money-back guarantee, and about 9% of first-year customers take the refund. We're building the referral flow this sprint and the engineer wants to just fire the reward email when the referred person creates an account, since that event already exists in our codebase. Tell me how to wire the payout.",
"expected_output": "A refusal of the signup-fired trigger, a verified value event chosen in its place, a holding period matched to the 60-day refund window, a pending-reward display, and the trigger ordering with the trade-off between delay and participation made explicit.",
"files": [],
"expectations": [
"Refuses the signup-fired trigger outright rather than accepting it as a cheap option to tighten later.",
"States that bare signup pays for tire-kickers and invites gaming.",
"Names a verified value event to fire on instead, such as first purchase or funded account, qualified activation, or a retention threshold.",
"Sets a holding period matched to the 60-day refund window so the reward clears only after refund exposure closes.",
"Promotes a retention-threshold or otherwise delayed trigger specifically because the refund exposure here is large.",
"Specifies displaying the reward as pending from day one, with wording that names the condition still to clear.",
"States that a visible pending reward is itself a motivator and costs nothing until the condition clears.",
"Gives the payout-trigger efficiency ordering explicitly (first purchase, then qualified activation, then closed-won or first invoice, then retention threshold, then staged payout) rather than only naming one recommendation.",
"States the trade-off that immediate rewards maximize participation and satisfaction while delayed rewards protect against refunds and churn.",
"Does not present bare signup as a ranked-but-lower option inside a menu of triggers."
]
},
{
"id": 5,
"prompt": "Marrowfield is a daily newsletter, 240,000 subscribers, free to read, monetized by ads at about $1.40 per subscriber per year. We want a referral ladder like the big newsletters: 3 referrals gets a sticker pack, 10 gets a tote, 25 gets a hoodie, 100 gets a weekend trip, and past that we'd keep inventing bigger prizes so our top 50 referrers keep pushing. Design the ladder.",
"expected_output": "A capped ladder: the open-ended top rung refused with the documented uncapped-ladder failure as justification, an entry micro-milestone below the 3-referral rung, a progress display tied to the goal-gradient effect, non-cash top rungs whose perceived value outruns real cost, and rung economics checked against the $1.40 per-subscriber annual value.",
"files": [],
"expectations": [
"Rejects the open-ended top of the ladder and requires a capped top tier with a named final rung.",
"Justifies the cap by citing the documented uncapped-ladder failure where escalating rewards reached free vehicles and the program was killed for adding too much cost.",
"Describes an uncapped ladder as an open-ended liability that grows with the most successful advocates.",
"Checks the fit test for tiering at all: high referral volume, plausible repeat referrers, and advocacy concentrated in a small group.",
"Keeps the first tier trivially achievable from the people immediately around a subscriber, and adds an entry micro-milestone below the 3-referral rung.",
"Requires a visible progress bar or equivalent progress display.",
"Names the goal-gradient effect as the behavioral mechanism the progress display exploits.",
"Requires perceived value to escalate faster than real cost, keeping top rungs non-cash (swag, access, status) rather than cash.",
"Raises the friend-side incentive modestly at higher tiers as well as the advocate's, because top referrers reach progressively colder audiences.",
"Checks the rung economics against the $1.40 annual value per subscriber, so cost per acquired subscriber stays a fraction of it.",
"Specifies a verified subscription as the qualifying event for a ladder credit, not a click or a share."
]
},
{
"id": 6,
"prompt": "Grantsloe is an $1,800/month B2B compliance platform sold through AEs, 76% gross margin, average contract 26 months. Our champions are compliance managers at mid-market banks. The plan is a $250 Visa gift card to the champion when their referral signs a contract. Legal hasn't looked at it yet. Write me the incentive structure.",
"expected_output": "A two-track structure replacing the gift card: company-level account value as the default track plus an optional capped non-cash individual token with an opt-out, both firing on closed-won or invoiced revenue, sized against B2B anchors, with the gift-policy and tax exposure named and the consumer-measured evidence base flagged.",
"files": [],
"expectations": [
"Flags the $250 gift card as colliding with corporate gift policies, noting gift cards are commonly treated as cash and that typical limits sit around $25-50 with logging requirements.",
"Warns that individual cash-equivalent rewards can pull the program into per-referrer tax reporting.",
"Recommends the two-track pattern: account-level value flowing to the customer's company as the default track, plus an optional individual track.",
"Makes the individual track modest, capped, non-cash, chooseable including a donate option, and carrying an explicit opt-out for referrers whose employer forbids gifts.",
"Names the principal-agent fork explicitly: the person referring is an employee who is not the party paying.",
"Sets the trigger for both tracks on closed-won or verified invoiced revenue, never on leads or trials.",
"Sizes the company-track reward against B2B anchors, using the 10-15% of LTV band or a percentage of CAC, rather than the 10-30% consumer band.",
"Computes a gross-profit ceiling from the $1,800/month price, 26-month contract length and 76% margin.",
"States that the evidence base behind this guidance is consumer-measured and B2B-inferred, and flags that caveat to the user.",
"Raises the higher B2B disclosure sensitivity: a champion recommending a vendor to their own network carries a material connection their employer may also police.",
"Notes the B2B scaling ceiling, that customer-referral programs migrate into affiliate or partner programs as contract value grows, and names the migration signal to watch for."
]
},
{
"id": 7,
"prompt": "Sabledrum is a $4/month consumer habit-tracker. We're 18 months old, 40,000 paying users, 91% gross margin, average tenure 5 months, CAC $11. Almost nobody mentions us to friends: our post-signup survey shows under 1% arrive via word of mouth, and our NPS is 14. The board wants a referral program in the plan for next quarter. Design the rewards.",
"expected_output": "A \"not yet\" recommendation rather than a design: the disqualifier check fails on both the absent organic word-of-mouth signal and an LTV too small to fund a reward above the motivation floor, with what would have to change first before revisiting.",
"files": [],
"expectations": [
"Concludes \"not yet\" and declines to produce a reward structure, rather than designing one anyway because it was asked for.",
"States the disqualifier reasoning, that a referral program captures existing word of mouth and does not create it, and cites the under-1% word-of-mouth share as evidence there is none here to capture.",
"Computes a gross-profit ceiling from $4/month, 5-month tenure and 91% margin (LTV around $20, gross profit around $18).",
"Shows that the resulting ceiling sits in single-digit dollars, below any meaningfully motivating reward.",
"Connects that to the non-monotonic floor: the only affordable reward here is a token one, and a token reward performs worse than no reward at all.",
"Recommends against launching rather than proposing a scaled-down program that fits inside the small ceiling.",
"Names what would have to change first (tenure, LTV, or the organic word-of-mouth share) before the question is worth revisiting.",
"Offers recognition-based advocacy or fixing the underlying retention problem as the alternative to a paid program.",
"Does not fall back to free months, in-kind product value, or a sweepstakes as a way to launch something anyway despite the failed economics.",
"States the general rule that when no structure clears the pass threshold, the recommendation is not to launch."
]
},
{
"id": 8,
"prompt": "We launched refer-a-friend at Corvidae, a $29/month B2C fitness app, four months ago: $15 credit to the referrer, $15 off the friend's first month, paid when the friend's first invoice clears. Traffic is fine, the terms page is fine, but only 0.9% of our active users have ever tapped 'share'. Our growth lead wants to raise both rewards to $25. Talk me out of it or tell me he's right.",
"expected_output": "A diagnosis that the low share rate is an ask problem, not a reward-size problem, with placement and timing fixes named, the reward raise argued against as a permanent budget increase against an hour of work, and the diagnostic order stated.",
"files": [],
"expectations": [
"Diagnoses the symptom as a problem with the timing, placement and friction of the ask, not with reward size.",
"States that placement of the ask moves participation more than reward size does.",
"Identifies the reward as not the variable at fault for this particular symptom, and says to fix the ask before touching it.",
"Advises against raising both rewards to $25 as the first move.",
"Contrasts the effort and reversibility: raising a published reward is a permanent budget increase that is hard to withdraw, while moving the prompt is about an hour of work.",
"Gives the diagnostic efficiency order explicitly, leading with repositioning the ask, then reweighting the split toward the friend, then raising the size.",
"Names concrete placement or timing changes rather than only advising to improve the ask.",
"Distinguishes this symptom from the different symptom where shares do happen but friends fail to convert, which would point at the friend-side split instead.",
"Applies the Rule of 100 to the $29 price, recommending percentage or free-month framing over \"$15 off\".",
"Does not propose an incrementality holdout as the first action for this symptom."
]
},
{
"id": 9,
"prompt": "Two things happening at Verdigris Home, a DTC cookware brand with a $180 average order and 54% gross margin. Referral-attributed orders are up 40% year over year and the referral dashboard says the program drove $2.1M last year. But blended CAC went from $61 to $88 in the same period and total new customers is basically flat. Our CMO wants to double the referral reward from $20 to $40 to push harder. What do you think?",
"expected_output": "A diagnosis that flat acquisition with rising CAC is the non-incrementality signature, promoting the holdout test ahead of every cheaper fix, rejecting the attributed-revenue number, and holding the reward increase until the holdout reads.",
"files": [],
"expectations": [
"Identifies the combination of flat total acquisition and rising blended CAC as the signature of a program that is paying for customers who would have come anyway.",
"Promotes the incrementality holdout immediately rather than leaving it last in the fix order where it normally sits.",
"States that this is the one symptom no cheaper fix explains, which is why the holdout jumps the queue.",
"Rejects the $2.1M attributed-revenue figure as the basis for judging the program.",
"Names incremental CAC and referred-cohort retention as the metrics to judge the program on instead.",
"Advises against doubling the reward from $20 to $40 before the holdout reads.",
"Specifies the holdout design: withhold the program from a slice of customers or a geography and compare total acquisition and CAC against the exposed group.",
"Warns that referral is the channel most prone to word-of-mouth leaking into the control group, so the holdout must be segmented strictly.",
"Sets an honest time cost for the holdout of roughly a quarter, including waiting out a full purchase cycle to read it.",
"Names the crude pause or on/off check only as a fallback for when the base cannot be segmented cleanly, noting seasonality confounds it.",
"Names the follow-on action if the holdout confirms low incrementality: narrow the reward to low-organic-reach segments rather than paying the whole base."
]
},
{
"id": 10,
"prompt": "Knollwick sells a one-time $2,400 home sauna kit at 41% gross margin. Buyers essentially never buy a second one. We also have an active owners' forum where about 6,000 people post build photos and answer each other's questions, and that's where most of our word of mouth happens. Marketing wants to give both sides a 10% discount code. Pick the rewards.",
"expected_output": "An asymmetric structure by reward type: a cash-equivalent to the referrer who will never rebuy, the price reduction to the friend framed as an absolute amount, the 10%/10% plan shown to breach the gross-profit ceiling, and the forum's in-group surfaced as a status option with the cash-versus-status tension resolved explicitly.",
"files": [],
"expectations": [
"Rejects giving the referrer a discount code, on the grounds that a discount on a product they will never buy again is worthless to them.",
"Recommends a cash-equivalent for the referrer and the price reduction for the friend, breaking the proposed symmetry deliberately.",
"Applies the Rule of 100 to the $2,400 price, framing the friend's reward as an absolute dollar amount rather than as \"10% off\".",
"Computes a gross-profit ceiling from the $2,400 price and 41% margin (gross profit around $984) and gives the resulting ceiling figure.",
"Shows that a 10% friend discount alone, around $240, consumes most or all of that ceiling before any referrer reward is added.",
"Surfaces the owners' forum as a real in-group that makes status, recognition or early access a live reward option at near-zero cost.",
"Warns that monetary rewards can crowd out image motivation where referring is a public, identity-driven act.",
"Resolves the tension between paying the referrer in cash-equivalents and preferring status in a public in-group explicitly, for example by offering the referrer a choice, rather than silently picking one.",
"Notes that status only works when a real in-group exists to signal to, and that fabricated exclusivity backfires.",
"Notes that the high price and long commitment raise the friend's conversion barrier, which argues for weight on the friend side."
]
},
{
"id": 11,
"prompt": "Lanternwake is a $600/month B2B SaaS. We want to pay a percentage of first-year contract value to people who send us deals, and we'd like them to also earn a slice of whatever the people they themselves recruit bring in. These are independent consultants in our space, not our customers. What percentage should we pay, and how do we structure the tiers?",
"expected_output": "A scope handoff: the payees are professional affiliates rather than end customers, so the contract-value commission and the cascading network payout both fall outside customer refer-a-friend incentive design, with the affiliate commission structure skill named as the correct destination.",
"files": [],
"expectations": [
"Identifies the payees as professional affiliates or partners rather than end customers, and says this falls outside customer refer-a-friend incentive design.",
"Routes the user to the affiliate commission structure skill, naming it as mbfinotti/partnerships-skills@affiliate-commission-structure.",
"Declines to produce a commission percentage for this partner motion inside this engagement.",
"Names the share-of-first-year-contract-value anchor as pricing a partner motion rather than as a customer-referral sizing anchor.",
"Names the cascading structure, commission on the referred network's own activity, as an affiliate or network structure that sits outside this design space.",
"States that customer referral and affiliate motions carry different economics and different tooling.",
"Notes that B2B customer-referral programs commonly migrate into affiliate or partner programs as contract value grows, which is what this request signals.",
"Does not present cascading commission as a tier option alongside flat, threshold ladder and stepwise shapes.",
"Keeps any in-scope end-customer refer-a-friend design the user may also want clearly separate from the consultant partner program."
]
}
],
"trigger_queries": [
{ "query": "How big should our refer-a-friend reward be?", "should_trigger": true },
{ "query": "Design the incentive structure for our customer referral program", "should_trigger": true },
{ "query": "should we give the reward to the person referring or the person who signs up?", "should_trigger": true },
{ "query": "give $20 get $20 — is that the right offer for us?", "should_trigger": true },
{ "query": "what's a good referral bonus for a $39/month SaaS", "should_trigger": true },
{ "query": "double-sided vs single-sided referral rewards, which performs better", "should_trigger": true },
{ "query": "we want to launch refer a friend next quarter, what do we actually pay people", "should_trigger": true },
{ "query": "Should our referral rewards be tiered? About 8% of customers refer and a handful refer repeatedly.", "should_trigger": true },
{ "query": "our referral program gives $10 credit and nobody uses it", "should_trigger": true },
{ "query": "how much of our LTV can we afford to give away per referral", "should_trigger": true },
{ "query": "when should the referral reward actually pay out", "should_trigger": true },
{ "query": "cash or account credit for referral rewards?", "should_trigger": true },
{ "query": "is it better to reward the friend or the referrer", "should_trigger": true },
{ "query": "we're a newsletter and want a milestone reward ladder for subscribers who bring people in", "should_trigger": true },
{ "query": "what do customers get for bringing us a new customer", "should_trigger": true },
{ "query": "our users tell their friends about us but we don't do anything for them — what should we give them?", "should_trigger": true },
{ "query": "a customer brought us three new accounts last month and asked what's in it for her. what do we offer?", "should_trigger": true },
{ "query": "we want to turn our word of mouth into something we can actually budget for", "should_trigger": true },
{ "query": "how do we thank people who send us business without it feeling cheap", "should_trigger": true },
{ "query": "should we give a free month or a discount when someone refers us", "should_trigger": true },
{ "query": "referral reward sizing against CAC", "should_trigger": true },
{ "query": "our CFO wants to know whether the refer-a-friend budget is justified", "should_trigger": true },
{ "query": "is $5 too small for a referral incentive", "should_trigger": true },
{ "query": "design a give-X-get-Y offer for our ecommerce store", "should_trigger": true },
{ "query": "what reward do we give B2B customers who refer us to other companies", "should_trigger": true },
{ "query": "our referral rewards fire on signup and we're paying out for junk accounts", "should_trigger": true },
{ "query": "we want to cap how much one person can earn from referring", "should_trigger": true },
{ "query": "should the friend be able to see that the referrer got paid?", "should_trigger": true },
{ "query": "how do we structure escalating rewards for repeat referrers", "should_trigger": true },
{ "query": "our top referrer has earned $4,000 in credits and finance is panicking", "should_trigger": true },
{ "query": "referral program economics — what's the ceiling on what we can pay", "should_trigger": true },
{ "query": "swag or cash for our referral ladder rungs", "should_trigger": true },
{ "query": "we're thinking of offering a charity donation option for referrals", "should_trigger": true },
{ "query": "what should the friend get versus the referrer", "should_trigger": true },
{ "query": "our referral program drove a lot of signups but our CAC went up", "should_trigger": true },
{ "query": "does raising our referral reward actually get us more referrals", "should_trigger": true },
{ "query": "how much do we pay customers to bring friends", "should_trigger": true },
{ "query": "Please advise on the reward architecture for a proposed customer advocacy referral scheme.", "should_trigger": true },
{ "query": "we sell a one-time purchase, so a discount code is useless to the referrer — what else can we give them", "should_trigger": true },
{ "query": "gift cards to B2B champions for referrals, is that going to be a problem", "should_trigger": true },
{ "query": "should referral rewards be expressed as a percentage or a dollar amount", "should_trigger": true },
{ "query": "our referral program pays out 500MB of storage — is product credit a good reward", "should_trigger": true },
{ "query": "we want to relaunch our referral offer with better numbers", "should_trigger": true },
{ "query": "what's the right split between the two sides of a referral offer", "should_trigger": true },
{ "query": "customers keep asking whether we have a program for people who recommend us. we don't. what would it look like?", "should_trigger": true },
{ "query": "should we hold the referral payout until the refund window closes", "should_trigger": true },
{ "query": "referral reward for a marketplace — do hosts and guests get the same amount", "should_trigger": true },
{ "query": "how do we know if our referral program is actually incremental", "should_trigger": true },
{ "query": "we want to reward referrals with early access instead of money", "should_trigger": true },
{ "query": "is a sweepstakes entry a reasonable referral reward", "should_trigger": true },
{ "query": "we spend $60 acquiring a customer through ads. our happiest users would probably bring us people for less. how do we set that up?", "should_trigger": true },
{ "query": "what payout trigger makes sense for referral rewards in a long B2B sales cycle", "should_trigger": true },
{ "query": "help me pick the refer-a-friend reward amounts for our subscription box", "should_trigger": true },
{ "query": "What commission rate should we pay our affiliates?", "should_trigger": false },
{ "query": "Set up a recurring revenue share for affiliate partners with a 12-month cap", "should_trigger": false },
{ "query": "Draft the terms and conditions affiliates accept when they join our program", "should_trigger": false },
{ "query": "Does this Instagram caption need an #ad disclosure to be compliant?", "should_trigger": false },
{ "query": "One affiliate has a suspicious conversion rate — is this cookie stuffing?", "should_trigger": false },
{ "query": "Write the welcome email sequence for newly approved affiliates", "should_trigger": false },
{ "query": "Audit this month's affiliate payout file before we send the money", "should_trigger": false },
{ "query": "What metrics belong on our affiliate performance dashboard?", "should_trigger": false },
{ "query": "Write a cold outreach email to recruit affiliates in the fitness niche", "should_trigger": false },
{ "query": "Rank these five candidate technology alliances by expected value", "should_trigger": false },
{ "query": "Write our deal registration policy for partner-sourced opportunities", "should_trigger": false },
{ "query": "Draft the creative brief for the influencer we just signed", "should_trigger": false },
{ "query": "Define the criteria we use to shortlist creators for this campaign", "should_trigger": false },
{ "query": "What KPIs should we set before this creator campaign launches?", "should_trigger": false },
{ "query": "Prepare me to negotiate usage rights and exclusivity with a YouTuber", "should_trigger": false },
{ "query": "Write the first DM pitching a partnership to this creator", "should_trigger": false },
{ "query": "Plan the co-marketing launch with our new integration partner", "should_trigger": false },
{ "query": "Two partners are fighting over the same account — write the rules of engagement", "should_trigger": false },
{ "query": "Design our B2B channel program from scratch: tiers, benefits, and which motions to run", "should_trigger": false },
{ "query": "Model the P&L of this one reseller relationship", "should_trigger": false },
{ "query": "Which partner categories should we add next and in what order?", "should_trigger": false },
{ "query": "Map every partner type we work with and score them", "should_trigger": false },
{ "query": "Sequence the certification tracks our partners need to sell on their own", "should_trigger": false },
{ "query": "Should we list on the AWS Marketplace or the Salesforce AppExchange first?", "should_trigger": false },
{ "query": "Build the quarterly scorecard we review each partner against", "should_trigger": false },
{ "query": "Redesign our silver/gold/platinum partner tiers and the promotion rules", "should_trigger": false },
{ "query": "How do I break into partnerships from a sales role?", "should_trigger": false },
{ "query": "Write the job description and interview loop for a channel manager", "should_trigger": false },
{ "query": "I'm starting partnership work at a new company — where do I begin?", "should_trigger": false },
{ "query": "Set velocity limits and cooldowns so people can't farm our referral program", "should_trigger": false },
{ "query": "Someone created 40 fake accounts to claim referral credits — what enforcement ladder do we publish?", "should_trigger": false },
{ "query": "Write the identity verification gate people must pass before a referral reward is released", "should_trigger": false },
{ "query": "Design the loyalty points program for our coffee shop app", "should_trigger": false },
{ "query": "Build an employee referral bonus program for engineering hires", "should_trigger": false },
{ "query": "What should we pay employees who refer candidates we end up hiring?", "should_trigger": false },
{ "query": "Set up the sales commission plan for our AEs next year", "should_trigger": false },
{ "query": "Design the SPIFF for our reps this quarter", "should_trigger": false },
{ "query": "Write our customer loyalty tier benefits — bronze, silver, gold", "should_trigger": false },
{ "query": "How should we price our new pro plan?", "should_trigger": false },
{ "query": "Calculate our blended CAC and LTV from this revenue data", "should_trigger": false },
{ "query": "Our NPS is 22 — how do we improve it?", "should_trigger": false },
{ "query": "Write the copy for our referral landing page and the share modal", "should_trigger": false },
{ "query": "Which referral software should we buy, or should we build it ourselves?", "should_trigger": false },
{ "query": "Build the email automation that sends the referral invite and the reminder", "should_trigger": false },
{ "query": "How do I get more Google reviews from happy customers?", "should_trigger": false },
{ "query": "Design a customer advisory board and pick who to invite", "should_trigger": false },
{ "query": "Set up a case study program with our best customers", "should_trigger": false },
{ "query": "What should our free trial length be?", "should_trigger": false },
{ "query": "Our churn is 6% monthly — how do we fix it?", "should_trigger": false },
{ "query": "Write the affiliate program page for our website", "should_trigger": false },
{ "query": "What's a fair finder's fee for a consultant who introduces us to a client?", "should_trigger": false },
{ "query": "Design a rewards program for our warehouse staff", "should_trigger": false },
{ "query": "Build a gamified points system to increase in-app engagement", "should_trigger": false }
]
}
references/examples.md›
# Worked Examples and Case Studies
Documented programs to pattern-match against, one full positive worked example, and the benchmark numbers with their reliability flagged.
## Canonical programs, decomposed by the four decision variables
| Program | Sidedness | Reward type | Trigger | Tier/cap | The lesson |
| -------------------------------------------- | ------------------------------------------------------------------------- | ---------------------------------------------------------------------- | ------------------------------------------------------- | ---------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Dropbox | Both, symmetric (500 MB each) | In-kind product value (storage) | Install + login + email verified - not bare signup | Total-earned cap (16/32 GB) that doubles as an upsell moment | Product-denominated rewards at near-zero marginal cost; ~1/3 of signups were already word-of-mouth before launch - the program captured existing WOM |
| PayPal | Both, symmetric ($20 → $10 → $5 → ended) | Cash | Signup + linked card + first transaction | Per-referrer annual cap (later) | Cash fits a money product; the designed _exit_ - stepping the reward down as organic growth took over - is the real lesson |
| Airbnb | Both, asymmetric; several times more for referring a host than a guest | Account credit (travel credit) | Friend's first _completed_ stay | Credit cap | Pay more for the scarce side of a marketplace; their A/B test found altruistic framing ("give your friend $25") beat selfish framing globally |
| Robinhood | Both | Lottery-shaped: ~99% of "free stock" grants worth ~$5, tail up to $200 | Account approved + funded | $1,500/year cap | A probability distribution lets the headline advertise the tail while expected cost stays ~$5 |
| Morning Brew | Referrer-only | Swag/status ladder | Verified subscription | 3 → 5 → 10 → 25 → 50 → 100 → 1,000; first tier reachable "from the coworkers around you" | Threshold ladders suit high-volume/low-value referrals; ~$0.25 per acquired subscriber at the sticker tier |
| Developer cloud hosts | Both, asymmetric (friend gets a large intro credit, referrer a small one) | Account credit | Friend spends a threshold **and** stays active ~90 days | Non-expiring referrer credit | Retention-gated triggers for usage-based products; note several strong PLG dev platforms run _no_ program - strong organic WOM can make one unnecessary |
| Tesla (negative example) | Both | Escalating prizes up to free vehicles | Referred purchase | **Uncapped ladder** - the failure | Killed repeatedly for "adding too much cost"; lavish rewards for a product with intense organic demand mostly subsidize purchases that would have happened anyway |
| B2B SaaS at scale (pattern, not one company) | Referrer-side commission | Revenue share or account credit | Closed-won / invoiced revenue | Contract-value based | Several well-known SaaS products migrated their customer-referral credit programs into affiliate/partner programs as ACV grew - the customer-referral model has a scaling ceiling in B2B |
## Worked example: $49/month self-serve B2B SaaS
Inputs: $49/month entry plan, 80% gross margin, ~14-month average lifetime (LTV ≈ $686), paid CAC $220, referrers are end users at small companies who usually control the card themselves (principal-agent risk low).
1. **Ceiling**: $686 × 0.80 × 25% allowed ≈ **$137 total per successful referral** - the boundary, not the target.
2. **Sidedness**: double-sided, tilted to the friend (the $49 commitment is the hard action). Test a friend-only arm against it.
3. **Type**: account credit both sides (subscription product, referrers rebuy monthly, credit costs margin only when redeemed).
4. **Size and framing**: friend gets 1 month free (framed "first month free" - on a sub-$100 price a whole free month reads bigger than "$49 off"); referrer gets $30 credit. Nominal total $79, real cost ≈ $63 per successful referral - well under the $137 ceiling, which is defined per successful referral. The CAC comparison is a different number: the pass threshold gates on cost per _incremental_ referred customer. About 50% of rewarded referrals would have converted anyway, so $63 ÷ 0.50 ≈ **$126 per incremental customer** against paid CAC $220 - a 1.7× advantage, not the 3.5× the raw $63 implies. Check it against blended CAC too, which sits below the paid figure and thins the margin further.
5. **Trigger**: referrer credit fires when the friend's _second_ paid invoice clears (survives the refund window and first-month churn); show it as pending from day one.
6. **Tiers**: none at launch - volume unproven. Revisit if >10% of referrers refer 3+.
7. **Caps**: 10 rewarded referrals per referrer per year; program budget reviewed quarterly against incremental CAC.
8. **Validation**: 3 arms (both-sides / friend-only / referrer-only) on 10% of the active base for 8 weeks; decision metric = new paying customers per 1,000 exposed accounts, checked against a no-program holdout.
## Benchmarks - with reliability flags
Peer-reviewed (durable):
- Referred customers: 16-25% higher LTV, ~18% lower churn (persistent), and 31-57% more onward referrals - the reason referral CAC may justifiably exceed naive parity with other channels.
- ~50% of rewarded referrals in academic data would have converted anyway (only the rest are incremental).
Vendor-published (directional only - undisclosed methodology, commercial incentive; never present these as facts):
- Share of customers who share: healthy programs 5-15%.
- Referred-visitor conversion: median 3-5% (e-commerce), top programs 8%+.
- Mature programs: 15-25% of new-customer acquisition.
- Minimum motivating reward: ≈ $20 or ~11% discount.
- The widely quoted "83% of satisfied customers are willing to refer, 29% do" traces to a single 2018 study no longer available online - cite with that caveat or not at all.
Reward amounts at named companies change frequently; verify against the live terms page before quoting one to the user.
references/reward-types.md›
# Reward Type Selection
How to choose what each party receives. The menu is stable; the choice is driven by product economics and referrer identity, not creativity.
## The menu, ranked by efficiency
Efficiency here is value delivered per unit of _effort to run the reward_ - setup work, ongoing administration, fulfillment burden, coordination with finance or legal, and how hard the reward is to withdraw once published. It is not value per dollar paid out; what each type costs you is a separate axis, and the two disagree.
- **efficiency**: in-kind product value > account credit == free months/seats > discount > status/recognition > cash/gift card > swag > charity > sweepstakes
- **cost to you** (cheapest first): in-kind == status > account credit == free months/seats > discount > swag > sweepstakes > cash == charity
- **value to the recipient** (strongest first): cash > account credit == free months/seats > discount > in-kind > swag > sweepstakes > status > charity
- **effort to run** (lightest first): discount > status > account credit == free months/seats > in-kind > cash > charity > swag > sweepstakes
- **compliance cost** (lightest first, only the three exposed types): cash/gift card > charity > sweepstakes. The other six trigger no review beyond the terms page every reward needs - but every published reward is hard to withdraw, so treat any headline number as a one-way door.
Two ties, both real:
- **Account credit == free months/seats**: the same billing primitive under two labels, the same margin-on-redemption cost, the same wiring. They differ only in how the number is framed: present a percentage under a $100 price and an absolute amount above it, never in what they cost or take to run.
- **In-kind == status on cost**: both grant something whose marginal cost rounds to zero. They split on effort and on whether an in-group exists to make status mean anything.
| Reward type | Real cost to you | Effort to run | Fits best when | Watch out for |
| -------------------------------------------------- | -------------------------------------------------- | -------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| In-kind product value (more storage, seats, usage) | Near-zero when marginal cost is low | A week: one provisioning hook, then near-zero | Freemium and infrastructure products where a unit of product is cheap to grant | Only motivates people already invested in the product |
| Account / product credit | Below face value (only costs margin when redeemed) | An hour to a week - billing systems already have a credit primitive | Subscription or repeat-purchase products; the referrer will buy again | Worthless to a referrer who will not purchase again; unredeemed credit still needs accounting treatment |
| Free months / seats | Deferred revenue cost | An hour to a week - same primitive as credit | Subscription products; symmetric "give a month, get a month" offers | Free-month hunters churn after redemption |
| Percentage or fixed discount | Margin give-up on the next purchase | Near-zero - a coupon code your commerce stack already issues | Retail/DTC with healthy margin; friend-side incentives | Attracts bargain hunters on the friend side if oversized |
| Status, recognition, early access | Near-zero | Near-zero to grant, a standing job to keep the in-group real | Communities, developer tools, prosumer products where referring is a public act | Needs a real in-group to signal to; fabricated exclusivity backfires |
| Cash / gift card | Full nominal cost, real P&L outflow | A quarter: payout rails, finance sign-off, then per-referrer tax reporting forever | Money-adjacent products (fintech, payments); one-time high-consideration purchases | Feels transactional; can crowd out the social motivation to recommend; triggers tax reporting thresholds in some jurisdictions |
| Physical swag (usually tiered) | Low unit cost + fulfillment overhead | A standing job: inventory, addresses, shipping, customs, replacements | Newsletters, community brands; milestone ladders | Logistics; works as a ladder, rarely as a single flat reward |
| Charity donation | Full cost, weaker conversion | A week to wire a donation vendor, then receipting every cycle | Mission-driven brands, image-sensitive audiences | Underperforms direct rewards on conversion in field tests - offer as a _choice_, never the only option; charitable-solicitation and cause-marketing rules vary by jurisdiction |
| Sweepstakes / lottery entry | Fixed prize pool | A quarter, most of it legal: official rules, registration where required, rules you cannot amend mid-run | High-volume, low-value-per-referral programs | Advertises a big number while expected payout stays small; some jurisdictions regulate lotteries |
**Default rung**: in-kind product value when the product has a cheap unit to grant, account credit everywhere else. Move to cash only when the referrer will never buy again or the product is money-adjacent - the one case where the efficiency leader is worth nothing to the person receiving it.
**What this order starves**: swag and status. Both lose every round on ratio because fulfillment and community upkeep are standing jobs, yet they are the only rewards that keep a high-volume, low-value-per-referral ladder affordable at its top rungs - a 1,000-referral tier is swag, never cash. Promote them when the reward is a ladder rung rather than a flat payout, and there is a real in-group to signal to.
This order is a default, not a law: it shifts with the product and with who has to run it. Re-rank against what you already know about the user before you present the menu:
- A product whose marginal unit costs near zero pushes in-kind further ahead of everything else.
- A finance team that refuses cash payouts deletes cash from the menu rather than demoting it.
- An existing rewards or fulfillment platform collapses the effort gap between swag, gift cards and credit, and re-ranks all three.
## Decision criteria, in order
1. **Can the reward be denominated in the product's own value?** If a unit of product (storage, credits, seats, usage) has near-zero marginal cost, in-kind rewards deliver high perceived value at low real cost, and redemption deepens engagement instead of draining cash - which is why it leads the efficiency order above. Account credit costs you the margin on redeemed credit, not the face value, a point most programs never model.
2. **Will the referrer buy again?** For one-time, high-consideration purchases, a discount on a product the referrer will never rebuy is worthless to them - pay the referrer in cash-equivalents and give the friend the discount. For subscriptions and repeat purchases, credit rewards the referrer _and_ drives their next purchase.
3. **Does reward-product congruence apply?** Field and lab evidence (Hu & Zhang, _Frontiers in Psychology_ 2021) shows utilitarian rewards work better for utilitarian products, hedonic rewards for hedonic products. A B2B tool is utilitarian: prefer cash, credit, or capability over gifts and experiences.
4. **Is referring a public, identity-driven act?** Where recommendations are visible (developer communities, open-source ecosystems, professional networks), image motivation does real work, and monetary rewards can crowd it out (Ariely, Bracha & Meier, _American Economic Review_ 2009 - evidence transferred from prosocial-behavior settings, treat as strong inference rather than direct proof). Status tiers, early access, and public recognition can outperform cash here at near-zero cost.
5. **Is the product innovative or novel?** Customers are _less_ likely to recommend innovative products when publicly rewarded for it - the reward taints the recommendation (Dose et al., _Journal of the Academy of Marketing Science_ 2019). The documented fixes: do not disclose the referrer's reward to the recipient, increase the reward size, or reward both sides.
## The B2B principal-agent fork
In B2B the person who refers is usually an employee, but the buyer is their company. Rewarding the individual with cash or gift cards collides with corporate gift policies (many prohibit cash outright and treat gift cards as cash; common limits are $25-50 per gift with logging requirements) and can pull the program into per-referrer tax reporting.
The durable pattern is **two-track**:
- **Company track**: account-level value - credits, discounts, service upgrades - flowing to the customer's company. Policy-neutral, procurement-friendly.
- **Individual track (optional)**: a modest, capped, non-cash token the individual can choose (or donate), with an explicit opt-out for referrers whose employer forbids gifts.
Fire both tracks only on verified invoiced revenue, never on leads or trials.
## Negative example - what not to do
A $200/month B2B analytics tool offers referrers a $25 Amazon gift card at friend signup. Three compounding errors: the reward is cash-equivalent to an employee (gift-policy collision), it fires on signup rather than a paid conversion (pays for tire-kickers), and $25 against a plausible four-figure LTV is deep in the too-small zone where a token payment performs worse than a plain "thank you" would. Better: 1 month of account credit to the company per closed-won referral, plus an optional charity-or-swag choice for the individual.
references/sidedness-and-triggers.md›
# Sidedness and Payout Triggers
Who gets rewarded, how the split works, and when the payout fires.
## Sidedness: the evidence, honestly stated
The founding taxonomy is **Reward-Me / Reward-You / Reward-Both** (Ryu & Feick, "A Penny for Your Thoughts", _Journal of Marketing_ 2007) - the academic origin of the industry's "sidedness" vocabulary. Its contingency findings still drive the decision:
- Rewards raise referral likelihood most for **weak ties and weaker brands** (strong ties to a strong brand refer without payment).
- For weak ties / weaker brands, rewarding the **referrer** matters most; for strong ties / strong brands, directing at least part of the reward to the **recipient** works better.
Three later results sharpen this:
| Finding | Evidence | Design implication |
| -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------- |
| Recipient-only rewards matched double-sided rewards costing **twice as much**; referrers were about equally likely to share either way, and the friend - who has the harder job (actually signing up and paying) - responds to their own incentive | Gershon, Cryder & John, _Journal of Marketing Research_ 2020, field experiments | Seriously test the friend-only arm; most programs over-invest in the referrer side |
| Prosocial splits (equal, or generous-to-friend) beat selfish (referrer-takes-all) schemes | Jain et al., _Journal of Management Information Systems_ 2021, randomized field experiment | If double-sided, split evenly or tilt toward the friend |
| Exception: referrer-side material rewards regain power when the referral **process is high-effort** | Baylor Keller Center synthesis of 8 studies | If sharing takes real work, pay the referrer for it - but first ask why sharing is that hard |
There is a genuine practitioner-vs-researcher tension worth surfacing to the user rather than hiding: prominent growth operators hold that the referrer's motivation is the primary lever, while the field-experimental evidence says the recipient's reward does more work. The honest recommendation is to run all three arms (referrer-only, friend-only, both) on a small share of the base - sidedness tests are cheap and this question is genuinely unsettled.
### The efficiency order
- **efficiency**: friend-only > double-sided > referrer-only
Effort is identical across the three arms - the same payout rail, the same terms page, one recipient or two - so the efficiency order collapses onto the value order in the table above, where recipient-only rewards matched double-sided programs costing twice as much.
**Default rung**: friend-only, headline written to the friend ("Give your friend X"), referrer payout kept out of that headline. Altruistic framing beat self-interested framing in a large marketplace's own global A/B test.
**Move to double-sided** when any of three conditions holds:
- The product is new or innovative: a visible referrer reward taints the recommendation, and rewarding both sides is one of the documented fixes.
- The brand is weak and the ties are weak (Ryu & Feick: the referrer's reward does the work there).
- You are running the three-arm test and want both-sides as the control.
**Move to referrer-only** only when the referral process is genuinely high-effort - a personal introduction, a demo the referrer sits in on. First ask why sharing is that hard; fixing the friction usually beats paying for it.
This ordering is a default, not a law. It rests on consumer-side field experiments, so downgrade your confidence for B2B. Re-rank it against anything you already know about this user's base - a brand nobody has heard of, or a referral flow that takes real work, moves the order before the evidence does.
## Asymmetric splits
Symmetric "Give $X, Get $X" is the default because it is easy to say and feels fair. The three tilts below are not competing options to rank: each fires on a different fact about the product, and at most one is true at a time. Break symmetry deliberately when the two sides have different jobs:
- **Tilt toward the friend** when the conversion barrier is high (expensive product, long commitment) - their reward subsidizes the hard action.
- **Tilt toward the referrer** when advocacy itself is costly (a personal introduction, a high-effort process), or when the referrer will never buy again (give them cash-equivalent, give the friend the discount).
- **Pay more for the scarcer side of a marketplace**: reward supply-side referrals (a new host, a new seller) several times more than demand-side ones.
## Disclosure
Whether the friend can see the referrer's reward changes behavior. Both options cost the same to implement and the evidence points both ways, so ranking them would be false precision - decide per context instead. Disclosing the referrer's reward can _increase_ referring by removing the hidden-agenda awkwardness (Wharton/Goethe reward-communication research), but for innovative products a _visible_ referrer reward taints the recommendation (Dose et al., _Journal of the Academy of Marketing Science_ 2019). Default: keep the referrer's payout out of the friend's headline; make the friend's benefit the visible message.
## Payout triggers
The trigger fires at the point where a referred signup becomes a real customer. Ranked by efficiency - protection bought per unit of engineering, instrumentation and cross-team coordination:
- **efficiency**: first purchase > qualified activation > closed-won/first invoice > retention threshold > staged payout
- **value** (protection against paying for a non-customer, strongest first): retention threshold > closed-won/first invoice > qualified activation > staged payout > first purchase
- **effort** (lightest first): first purchase > qualified activation == closed-won/first invoice > retention threshold > staged payout
The tie is real: qualified activation and closed-won each wire one new definition onto an event the business already emits - an activation event, or a CRM stage plus finance confirmation - with no holding state and no partial payouts either way.
**Bare signup is deleted from this menu, not ranked last.** It pays for tire-kickers, invites gaming, and no economics rescue it; a program firing on signup is a program to diagnose, not a design to choose.
| Trigger | Effort to run | Typical use | Trade-off |
| --------------------------------------------------------------------- | ---------------------------------------------------------------------------- | --------------------------- | -------------------------------------------------------------------- |
| First purchase / funded account | Near-zero - the billing event already fires | B2C, fintech | Standard; still exposed to refunds |
| Qualified activation (installed + verified + first meaningful action) | A week - define and instrument the activation event | Freemium, PLG | Filters tire-kickers at the cost of some legitimate payouts |
| Closed-won / first paid invoice | A week - a CRM stage plus finance confirmation, across two teams | B2B sales-led | The B2B norm; fire on verified invoiced revenue, not leads or trials |
| Retention threshold (spend ≥ $X **and** active N days) | A quarter - usage instrumentation, a holding state, a pending-reward display | Infrastructure, usage-based | Cleanest signal; delays gratification - show the pending reward |
| Staged payout (small at qualified lead, larger at conversion) | A quarter to build, then a standing reconciliation job | B2B long sales cycles | Keeps referrers engaged through a months-long cycle |
**Default rung**: first purchase / funded account. Move up to qualified activation when free signups are cheap to fake or the first purchase is trivially small.
**What this order starves**: the retention threshold - highest value, high effort, so it loses every round on ratio. Promote it whenever refund and churn exposure is large: usage-based billing, a generous refund window, or a first month that a meaningful share of customers cancel inside.
The ordering shifts with the business and with who builds it:
- A sales-led B2B company has no self-serve purchase event at all, which promotes closed-won to the top.
- A months-long cycle in which referrers go quiet promotes the staged payout.
- A team with no analytics engineer cannot reach the retention threshold at any price, which deletes it rather than demoting it.
Two design notes:
- **A visible pending reward is itself a motivator.** "You'll receive $50 when Alex completes their first month" keeps the referrer engaged and costs nothing until the condition clears.
- **Delay is also the margin guard.** Immediate rewards maximize participation and satisfaction; conditioned/delayed rewards protect against refunds and churn. Match the holding period to the product's refund window - that decision belongs to incentive design even though enforcement against deliberate abuse does not (that is a separate fraud-prevention concern).
## Caps
Every structure needs explicit caps before launch, because they are painful to introduce later:
- **Per-referrer annual cap**: bounds P&L exposure.
- **Total-earned cap**, where rewards accumulate: a storage or credit ceiling can double as an upsell moment, since the power referrer who hits the cap is a warm expansion lead.
- **Program budget ceiling**: reviewed against incremental CAC rather than attributed revenue.
references/sizing-and-tiers.md›
# Reward Sizing, Economics, and Tier Design
How much to pay, how to frame the number, how to know the spend is incremental, and how to build a ladder that does not blow up.
## The ceiling: compute it first
No consensus formula recommends an amount, but a family of anchors converges on one discipline: **total reward cost per successful referral must stay under the gross profit the referred customer creates, and referral CAC must undercut the next-best channel.**
```
Reward ceiling = referred LTV x gross margin x allowed acquisition %
```
Example: LTV $1,200, margin 70%, willing to spend 20% of gross profit on acquisition → ceiling = $168. That is an upper boundary, not a recommendation - set the actual reward well below it.
Count the _total_ cost against the ceiling: referrer reward + friend reward + fulfillment + the margin impact of any discount. And remember credit costs margin only when redeemed, not face value.
## The anchors: pick a target inside the ceiling
Three anchors, ranked by efficiency - confidence in the number per unit of analysis work it takes to compute:
- **efficiency**: % of CAC > % of LTV > payback discipline
- **value** (tightness of the bound it puts on the P&L, strongest first): payback discipline > % of LTV > % of CAC
- **effort** (lightest first): % of CAC > % of LTV > payback discipline
| Anchor | Published range | Effort to compute | Notes |
| ------------------------------- | ----------------------------------------------------------- | --------------------------------------------- | --------------------------------------------------------------------------- |
| % of CAC | 40-60% of blended/paid CAC | Near-zero - finance already reports it | Guarantees referral CAC beats the paid channel it displaces |
| % of LTV (combined, both sides) | 10-30% of first-year LTV consumer; 10-15% B2B | An hour, or a week when churn data is thin | The most common practitioner band |
| Payback discipline | Reward paid back by referred gross profit within ~12 months | A week - requires a cohort gross-profit curve | Compute against first 3/6/12 months of gross profit depending on churn risk |
**Default rung**: % of CAC. Move up to payback discipline when churn is front-loaded or the LTV figure is a guess dressed as a number - that is exactly the case the least-effort anchor cannot catch.
**What this order starves**: payback discipline, the strongest bound and the slowest to compute. It also swings hardest with who runs it: a team with a working cohort model reaches it in an afternoon, which re-ranks it first.
The commission-style anchor (a share of first-year contract value) is deleted rather than listed last - it prices a partner motion, which sits outside this skill. Route that case to `mbfinotti/partnerships-skills@affiliate-commission-structure`.
These bands are practitioner-published, not peer-reviewed - use them as starting points for a test, not as answers.
## The floor: the response curve is non-monotonic
The foundational result (Gneezy & Rustichini, "Pay Enough or Don't Pay at All", _Quarterly Journal of Economics_ 2000): introducing a _small_ payment produces worse performance than no payment at all, because it converts an intrinsically motivated favor into a poorly paid job. Applied here: a token reward against a meaningful purchase reads as insulting and can suppress referrals below the unpaid baseline.
Above the floor, returns diminish: past a moderate amount, social friction - not reward size - becomes the binding constraint, and doubling the reward moves the referral rate only marginally. So the shape is: **pay nothing (and lean on recognition), or pay meaningfully; never pay a little.** Fix friction and ask-timing before raising the number - placement of the ask moves participation more than size does.
Also justified by evidence when paying near the top of the range: referred customers are worth more. The strongest study in the field (Schmitt, Skiera & Van den Bulte, _Journal of Marketing_ 2011 - ~10,000 bank customers over 33 months) found referred customers churn ~18% less (persistently) and carry a 16-25% LTV premium. A €25 reward against a ~€40 value difference returned ~60% over six years.
Referred customers also go on to make 31-57% more referrals themselves (Gershon & Jiang, _Journal of Marketing Research_ 2024). Both effects vary by segment - the authors explicitly recommend selective rather than blanket rewards.
## Framing: the Rule of 100
Present whichever number _looks_ bigger:
- Price under $100 → frame as a **percentage** ("50% off" beats "$15 off" on a $30 plan).
- Price over $100 → frame as an **absolute amount** ("$200 off" beats "10% off" on a $2,000 contract).
At exactly $100 they are equal. Grounded in reference-dependence (people evaluate the number itself, not just what it denotes).
## Incrementality: the honest measure of the program
Roughly half of referral rewards in academic data go to customers who would have joined anyway (Kumar, Petersen & Leone's type-one vs type-two referral distinction, _Journal of Marketing_ 2010). The canonical freemium success story confirms the pattern: about a third of its signups already came from word-of-mouth _before_ its referral program launched - the program captured existing word-of-mouth rather than creating it.
Two diagnostics to build into the spec - **efficiency: holdout > on/off check**. Pausing the program costs no engineering but buys a read that seasonality and word-of-mouth spillover both confound; the holdout costs a week to design and actually answers the question. Fall back to the on/off check only when the base cannot be segmented cleanly.
1. **Holdout test**: withhold the program from a slice of customers or a geography. If total acquisition stays flat while CAC rises, rewards are cannibalizing organic referrals. Design the holdout strictly - referral is the channel most prone to word-of-mouth leaking into the control group.
2. **The crude on/off check**: pause the program briefly; if new-customer volume barely moves, most rewarded referrals were type-two.
Judge the program on **incremental CAC and referred-cohort retention**, never on the attributed-revenue dashboard.
## Tier design
Tier only when the fit conditions hold: high referral volume, plausible repeat referrers, and advocacy concentrated in a small group. A flat reward is simpler to explain, support, and account for - complexity kills participation.
Three shapes, ranked by efficiency - referrals unlocked per unit of build and upkeep:
- **efficiency**: flat > threshold ladder > stepwise
- **value** (extra referrals unlocked, strongest first): threshold ladder > stepwise > flat
- **effort** (lightest first): flat > stepwise > threshold ladder
The ladder outranks stepwise on efficiency despite costing more to build, because it is the only shape that raises output from a referrer you already have; stepwise merely re-times a payout the flat design would have made anyway.
| Shape | Effort to run | Mechanism | Fits |
| ---------------- | --------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- | ----------------------------------------- |
| Flat | Near-zero - one reward, one rule to explain | Same reward per referral for everyone | Default; low volume; B2B |
| Threshold ladder | A quarter to build (milestone accounting, progress display, cap enforcement), plus a standing fulfillment job if the rungs are physical | Escalating rewards at cumulative milestones (3 → 5 → 10 → 25...) | Newsletters, communities, high-volume B2C |
| Stepwise | A week - one instrumented event per step, plus partial-payout accounting | Different rewards for different actions (share, signup, first purchase) | Multi-step funnels with long conversion |
**Default rung**: flat. The ladder loses every round on ratio and is still the right answer for a high-volume, low-value-per-referral program - the condition that promotes it is the three-part fit test above, never the ratio. Re-rank against what the user already owns: a rewards platform that ships milestone tracking and progress bars out of the box erases most of the ladder's effort, and with it the gap to flat.
The cascading shape - commission on the referred network's own activity - is deleted from this menu rather than ranked last: it is an affiliate/network structure, and a user who needs it is designing a partner program. Route them to `mbfinotti/partnerships-skills@affiliate-commission-structure`.
Ladder rules, each traceable to a documented program outcome:
1. **Make the first tier trivially achievable** - low enough that a motivated customer can clear it from the people immediately around them. The hardest conversion is from zero referrals to one; an entry micro-milestone ("share once, get a small unlock") converts intention into action.
2. **Escalate perceived value faster than real cost.** Physical or in-product rewards with high perceived value and low unit cost (swag, exclusive access, product capacity) make ladders affordable; pure-cash ladders scale cost linearly with success.
3. **Show a progress bar.** The goal-gradient effect - people accelerate as they approach a goal - is the main behavioral lever of a ladder, and it only works if progress is visible.
4. **Cap the top tier, always.** The best-documented failure in this discipline is an automaker's uncapped referral ladder whose "secret levels" escalated to free cars, was killed for "adding too much cost", relaunched cheaper, and eventually shut down for good. An uncapped ladder is an open-ended liability that grows with your most successful advocates.
5. **Raise the friend's incentive modestly at higher tiers too**, not only the advocate's - top referrers reach progressively colder audiences who need more reason to act.
## Budget and P&L notes
State in the spec who owns the cost line and how it is booked:
- Practice at scale treats referrer rewards as customer-acquisition cost, a sales & marketing expense.
- Any payment beyond the fair value of the referral service counts as a reduction of revenue.
- Friend-side discounts are margin give-up.
- Unredeemed credit is a liability until it expires.
Getting finance sign-off on this before launch is cheaper than re-architecting the program after the first audit question.
SKILL.md›
---
name: referral-incentive-design
description: Design the reward structure of a customer refer-a-friend program - who gets rewarded (referrer, friend, or both), reward type (cash, account credit, discount, product value, status), how much to pay against LTV/CAC and gross margin, when the payout fires, and whether to build a tiered reward ladder. Use whenever the user mentions referral incentives, refer-a-friend rewards, referral bonus amounts, give-X-get-Y offers, double-sided referral rewards, or asks how big a referral reward should be, even if they never say incentive. Covers end-customer rewards only. Do NOT use for commission rates paid to professional affiliates - use mbfinotti/partnerships-skills@affiliate-commission-structure instead.
license: MIT
metadata:
author: Maya-Beth Finotti
version: "1.0.9"
---
# Referral Incentive Design
Design the reward structure of a customer refer-a-friend program. No canonical named framework exists for this discipline - referral is one letter inside measurement funnels like AARRR, not a design method. What exists is four decision variables, each with its own evidence base, worked through in order because each shapes the next:
1. **Sidedness** - reward the referrer, the friend, or both (the Reward-Me / Reward-You / Reward-Both taxonomy, Ryu & Feick, _Journal of Marketing_ 2007).
2. **Reward type** - cash, account credit, discount, product value, status, or donation.
3. **Reward size** - anchored to gross margin, LTV, and CAC, with a floor set by non-monotonic response (a too-small reward performs worse than none).
4. **Payout trigger and tier structure** - when the reward fires, and whether rewards escalate with cumulative referrals.
## Interview
Ask before recommending anything. One question per message; offer multiple-choice answers when possible. Skip a question only when the user already gave the answer.
- Is the business B2B or B2C? Subscription or one-time purchase? Self-serve or sales-led?
- What are the gross margin, average LTV (or first-year revenue), blended/paid CAC, and entry-plan price?
- Do customers already recommend the product organically? Roughly what share of new customers arrive that way?
- Will the referrer plausibly buy again (account credit has value to them), or not (only cash-equivalents do)?
- For B2B: is the likely referrer the buyer, or an employee whose company pays?
- By what date must this show results? A hard quarter-end date promotes the fast rungs - a flat reward on a first-purchase trigger - and rules out ladders and retention-gated triggers, which need months to read.
- One-off acquisition win, or a compounding asset? Compounding promotes in-kind and status rewards, a tier ladder, and a retention-gated trigger; a one-off win promotes a flat friend-side offer.
- What is the effort ceiling: who builds and administers this, how many hours do they have, and can a published reward be withdrawn later? A low ceiling deletes swag, sweepstakes, and staged payouts outright; a dedicated owner puts the ladder back on the table.
## Workflow
1. Run the Interview above; collect every answer the structure depends on before designing.
2. Check the disqualifiers before designing. A referral program captures existing word-of-mouth; it does not create it. No organic referral signal, or an LTV too low to fund a meaningful reward, means the honest answer is "not yet" - say so and stop. Confirm the question is about end-customer referrals, not external affiliates or channel partners (different economics, different tooling; B2B programs tend to migrate from the former to the latter as ACV grows).
3. Decide sidedness, working from [references/sidedness-and-triggers.md](references/sidedness-and-triggers.md). Efficiency order: **friend-only > double-sided > referrer-only** - effort is identical across the three arms, so the value evidence decides, and field experiments found recipient-only rewards matching double-sided rewards that cost twice as much. Lead with friend-only; move to double-sided for an innovative product or a weak brand with weak ties, and to referrer-only only for a genuinely high-effort referral process.
4. Choose the reward type per side from [references/reward-types.md](references/reward-types.md), leading with its efficiency order: **in-kind product value > account credit == free months > discount > status > cash > swag > charity > sweepstakes**. Then apply the criteria that override it:
- Cash-equivalents when the referrer will never rebuy or the product is money-adjacent.
- Status over cash when referring is a public, identity-driven act.
- Charity only ever as a choice.
For B2B, resolve the principal-agent fork (individual reward vs company account credit) explicitly.
5. Size the reward using [references/sizing-and-tiers.md](references/sizing-and-tiers.md): compute the gross-profit ceiling, pick a target inside it starting from the least-effort anchor (**% of CAC > % of LTV > payback discipline**), verify the floor (non-monotonic response: pay meaningfully or pay nothing), and frame the number with the Rule of 100 (percentage below $100, absolute amount above).
6. Set the payout trigger on a verified value event, never bare signup. Efficiency order: **first purchase > qualified activation > closed-won/first invoice > retention threshold > staged payout**; promote closed-won for a sales-led B2B motion, and the retention threshold when refund or churn exposure is large. Match the holding period to refund risk and sales-cycle length.
7. Decide whether to tier, against the ladder rules in [references/sizing-and-tiers.md](references/sizing-and-tiers.md). Efficiency order: **flat > threshold ladder > stepwise**; the ladder is worth its build cost only when volume is high, repeat referrers are plausible, and advocacy concentrates in a few advocates. First tier trivially achievable; top tier always capped - the best-documented failure in this field is an uncapped ladder.
8. Present the four decisions one by one - options considered, trade-offs, and your recommendation - and validate each with the user; revise on pushback before moving to the next.
9. After explicit approval on all four, emit the full structure as a testable spec (next section), grounded in a matching worked example from [references/examples.md](references/examples.md) when one fits the user's situation.
10. If your harness has persistent memory, memorize the approved spec's key decisions and economics - a later diagnosis run starts from them instead of re-interviewing.
Every ordering above ranks value returned per unit of _effort_ - build time, administration, fulfillment, coordination, and how hard the choice is to reverse once published. Never rank by payout size; what a reward costs is a separate axis, and the two disagree often enough that a blended rank hides the case where the cheapest option and the strongest one are different options.
Each order is a default, not a law: it shifts with the situation and with who has to execute it. Re-rank against what you already know about this user before presenting any menu:
- A product whose marginal unit costs near zero promotes in-kind rewards above everything.
- A finance team that refuses cash payouts deletes cash from the menu rather than demoting it.
- An existing rewards platform erases the effort gap that puts ladders and swag near the bottom.
Say out loud which fact moved which option, so the user can argue with the reasoning instead of the row order.
## The Reward Structure Spec
Deliver every engagement as this artifact - a decision record the user can hand to product, finance, and a test-plan owner:
```
REWARD STRUCTURE - <product>, <date>
Sidedness : both | friend-only | referrer-only, split, disclosure choice
Reward (friend) : type, amount, framing ("first month free", not "$49 off")
Reward (referrer): type, amount, framing
Payout trigger : verified event per side + holding period + pending-state display
Tiers : ladder breakpoints + top-tier cap, or "flat - revisit when <condition>"
Caps : per-referrer/year, total-earned, program budget ceiling
Economics : gross-profit ceiling, projected cost per referred customer vs current CAC
Validation : test arms, % of base exposed, duration, decision metric, incrementality check
KPIs : participation rate, share-to-signup conversion, cost per incremental referred customer vs CAC
```
Three rules about the spec itself:
- **Every line carries a reason.** "Referrer credit fires on the second invoice - survives the refund window" is reviewable; a bare amount is not.
- **The validation plan is not optional.** Sidedness is genuinely contested between practitioners and field research, so name its test arms (both / friend-only / referrer-only, on 5-10% of the base) and its incrementality check. The incrementality check means holding out a slice: flat acquisition volume with rising CAC shows the program pays for customers who would have come anyway.
- **The spec has a pass threshold.** Projected cost per incremental referred customer must land below current blended CAC and inside the gross-profit ceiling. Iterate size, type, or trigger until both hold; if no structure passes, recommend not launching.
## Diagnosing an Existing Program
When the user arrives with a running program instead of a blank page ("we give $10 credit and nobody uses it"), work the fixes in efficiency order - most program repaired per unit of effort, which is not the same as the cheapest fix first:
**efficiency**: reposition the ask > reweight the split toward the friend > raise the size above the floor > cap the ladder and move the trigger > switch reward type > run the incrementality holdout
| Symptom | Check first | Effort | Variable at fault |
| ------------------------------------------- | ----------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------- |
| Few customers ever share | Timing and friction of the ask - placement moves participation more than reward size does | An hour - move the prompt, change the copy | None - fix the ask before touching the reward |
| Shares happen, friends do not convert | Friend-side offer strength - the friend has the harder job | An hour - reweight a budget you are already spending | Sidedness/split |
| Participation dropped after adding a reward | Token amount below the motivation floor turned a favor into a badly paid job | An hour of config, against a permanent budget increase | Size |
| Power referrers milk the ladder | Missing top-tier cap or signup-fired trigger | A week - and capping a published ladder is the hardest change here to make stick | Tiers/trigger |
| B2B referrers decline rewards | Corporate gift policy collision | A quarter - finance, procurement and legal all touch it | Type: switch to the two-track pattern (company credit + optional individual token) |
| Healthy referral volume, rising CAC | Incrementality - rewards paying for organic word-of-mouth | A quarter - design the holdout, wait out a full cycle to read it | Economics: run the holdout, then narrow the reward to low-organic-reach segments |
The order starves the incrementality holdout: it lands last on ratio and first on truth, because it is the only check that can reveal the whole program is buying customers who were coming anyway. Promote it immediately when referral volume looks healthy while CAC climbs - that is the one symptom no cheap fix explains. Re-rank the rest against the user's constraints too: a team that cannot ship UI this quarter cannot reposition the ask, which pushes the split reweight to the top.
## B2B and B2C
Identical in both, whether B2B or B2C:
- The four decisions and their order.
- The gross-profit ceiling.
- The non-monotonic floor.
- The Rule of 100 framing.
- The payout trigger on a verified value event.
- The incrementality haircut.
Do not reinvent any of them per side.
The evidence base is the asymmetry worth stating out loud: the field experiments and academic results this skill leans on are overwhelmingly consumer-side. Treat every number here as B2C-measured and B2B-inferred, and say so when advising a B2B program.
| Dimension | B2C | B2B |
| ---------------------- | --------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Who is rewarded | The customer, who is also the payer | Often an employee who is not the payer - resolve the principal-agent fork explicitly at step 4 |
| Reward vehicle | Account credit or cash-equivalent to the individual | Company account credit by default; an individual reward is a gift-policy and procurement question before it is an incentive |
| Payout trigger | First paid invoice or an activation milestone | Later - closed-won or a retention threshold, because the cycle is longer and refund exposure larger |
| Scaling ceiling | Holds at scale | Real: as ACV grows, customer-referral programs migrate into affiliate or partner programs. Watch for the migration signal rather than stretching this structure past it |
| Disclosure sensitivity | Moderate | Higher - a referrer recommending a vendor to their own network carries a material connection their employer may also police |
## Invocation Examples
- "Design the refer-a-friend rewards for a $49/mo B2B SaaS - 80% margin, 14-month average tenure, $220 paid CAC."
- "We give $10 credit both sides and nobody uses it. Diagnose which of the four variables is at fault."
- "Should our referral rewards be tiered? About 8% of customers refer, and a handful refer repeatedly."
## Reference
- `mbfinotti/partnerships-skills@affiliate-commission-structure` for commission design aimed at professional affiliates and partners.
- `mbfinotti/partnerships-skills@referral-abuse-guardrails` for the enforcement layer around caps - velocity limits, cooldowns, burst detection, verification gates, and the published wording that makes a cap stick. This skill sets the cap values; that skill makes them hold.