Referral And Affiliate Programs . Agency OS

What it does

Produces a complete referral or affiliate programme: the trigger moments to ask at, the share mechanism, the incentive structure, and the reward ceiling computed from your own LTV and margin rather than from someone else's benchmark. On the affiliate side it adds commission structure, attribution window, recruitment sources and an enablement kit. It also writes the launch announcement, the nurture cadence, the fraud rules and terms, the measurement set and a launch checklist. In client mode the whole programme is built in the client's voice and economics as a deliverable you can charge for.

Say this to start

This skill has no button. You start it by saying what you want. Any of these will do it:

> referral program

> affiliate program

> get more referrals

> refer a friend

> partner program

> referral incentive

When to reach for it

most of your work already arrives by word of mouth with no system behind it
a client asked for a refer-a-friend or ambassador programme
you need to know what a referral reward can actually cost you
an existing programme is live and almost nobody refers
referrals arrive but they do not convert
recruiting affiliates, creators or partner agencies

When NOT to use it

If you actually want	Use this instead
a client who is at risk, has gone quiet, or has given notice, including the single referral ask made at a graceful exit	`client-retention`
cold outbound prospecting to strangers	`agency-sales-system`
qualifying and closing a referred lead once it lands	`sales-conversion-engine`
building the referral landing page and the CRM automations behind it	`funnel-architect`
designing the incentive into the offer economics when LTV and margin are not defined yet	`offer-architect`

Before you start

What you need	Why
`offer-output.md` containing customer LTV, gross margin, price point and target CAC	the maximum reward is calculated as LTV times gross margin minus target CAC, so without these numbers the incentive cannot be sized and the skill sends you to offer-architect	Required
Brand DNA complete, or these supplied by hand: business name, product or service, price point, average customer LTV, current CAC and target audience	the skill will not proceed on no context at all; it asks for Build 1 or for this list	Required
A decision on which workspace you are in, `_default` for your agency or `client--*` for a client deliverable	client-mode output carries the client's business name, voice and economics, never the agency's	Required
`verified-claims.md` populated with consented proof	any result, testimonial or named client that appears in invite emails, landing copy or an affiliate enablement kit must trace to the registry or it comes back generic	Required
`ica-output.md`	used to profile who the highest-value referrers are and what motivates them, otherwise you are asked directly	Optional
A referral or affiliate tool chosen, or a budget line for one	the skill designs the programme and its terms but does not stand up tracking links, payouts or the landing page	Optional
Current programme data if one already exists: referral rate, incentives already tried	the optimisation path depends on knowing which part of the loop is broken, awareness, share rate or conversion	Optional

How it runs

Workspace and modeDetects _default for your own agency or client--* for a client deliverable, then loads brand memory under the freshness rules and applies any learnings.md patterns tagged referrals that have recurred three or more times.
Pull economics before askingReads LTV, margin, price point and audience profile out of offer-output.md and ica-output.md first. You are only asked for what the workspace cannot answer: programme type, B2B or B2C, current state, product shareability, tooling and incentive budget.
Choose referral or affiliateSeparates a customer referral programme (referrer is an existing customer, one-time reward, higher trust, lower volume) from an affiliate programme (ongoing commission, higher volume, variable trust). For a service business it will usually steer you toward referral partners over affiliates.
Design the loopSets the trigger moments to ask at (first delivered win, milestone hit, glowing review, renewal or upgrade), picks the share mechanism from the ranked list, and chooses between single-sided, double-sided and tiered incentive structures with the trade-off of each stated.
Size the incentive from your own numbersComputes the ceiling as (Customer LTV times Gross Margin) minus Target CAC and presents your math, not a benchmark range. Incentive type is then matched to the business: credit, free months, scope opens, cash, gifts or charity donation.
Affiliate track and fraud rulesIf affiliates are in scope, sets commission structure, attribution window, recruitment sources and the enablement kit. Then writes the fraud controls (verification, delayed payout until a meaningful action, per-period caps, clawback on refunds, manual review) and the programme terms, before launch rather than after.
Copy and quality gateWrites the launch announcement and the nurture cadence at day 7, day 30, day 60 and after every customer milestone. Each customer-facing asset runs the two-tier claim check until it passes, then the critique skill scores the launch announcement and landing copy, flagging anything below 70 for revision.
Launch checklist and measurementDelivers the before-launch, launch and first-30-days checklists alongside the measurement set: active referrers, referral conversion rate, rewards paid, share of new customers from referrals, referral CAC against other channels, and programme ROI. Files save to [target]/growth/referral-program/ and get logged to assets.md.

What you get

A programme design document: trigger moments, share mechanism, incentive structure and the reward ceiling with the math shown
Affiliate commission structure and attribution window where affiliates are in scope
An affiliate enablement kit spec: tracking links, product overview, audience description, competitor comparison, creative assets, sample copy, cleared case studies, demo access, FAQ and payment terms
Fraud prevention rules and programme terms and conditions
A launch announcement email and the nurture cadence at day 7, 30 and 60 plus milestone prompts
An optimisation playbook mapping each failure (low awareness, low share rate, low conversion, fraud, one-time referrers) to its fix, plus the A/B tests worth running
The measurement set including viral coefficient, referral rate and programme ROI formulas
Launch checklist covering before launch, launch day and the first 30 days
Every deliverable as .md and .docx in [target]/growth/referral-program/, named [YYYY-MM-DD]-referral-program-[deliverable-slug]
An assets.md entry tagged referrals

Honest limits

Read this before you rely on it

It designs the programme, it does not build it. Tracking links, the referral landing page, payouts and CRM automations sit with funnel-architect and whichever referral tool you choose.
Every benchmark inside the skill is internal calibration only: reward ranges, commission percentages, attribution windows and what counts as a good referral rate. None of it may appear in a client deliverable, which must use the client's own numbers or carry a proof gap flag.
The reward ceiling is only as trustworthy as the LTV, margin and CAC figures you feed it. A wrong LTV produces a reward that quietly eats the programme's profit.
Copy goes through the claim gate and then critique scoring, with anything under 70 sent back. Budget for a revision pass rather than expecting first-draft-final.
Fraud rules get written but you have to enforce them. Verification, delayed payout, caps, clawbacks and manual review of suspicious patterns are ongoing operational work, not a one-time setup.

Where people go wrong

The mistake	Do this instead
Copying an incentive amount from a programme you admire	Compute your own ceiling: LTV times gross margin minus target CAC. Anything above that line is bought revenue at a loss.
Asking for referrals on a schedule or in a monthly newsletter	Ask at high-intent moments: right after the first delivered win, after a milestone, after exceptional support or a glowing review, and after a renewal or upgrade.
Launching first and writing terms and fraud rules later	Write them before launch. Retrofitting rules onto a live programme is expensive and public.
Putting all the effort into the referrer and none into the person referred	The referred user's landing experience is what converts the programme. Make the referrer's endorsement visible on that page.
Defaulting to a single-sided reward on a low-ticket offer	Test double-sided first. Single-sided suits high-value products where the introduction itself is the gift.
Quoting an industry referral benchmark in a client report	Use the client's own baseline data, or mark a proof gap and go and get the number.

Worth knowing

For a service business or agency, 'referral partner' usually beats 'affiliate'. Complementary providers such as web designers, bookkeepers and consultants send warm introductions for a flat fee or a reciprocal referral, which converts better than a commission link and needs no affiliate tooling to start.