referral-program

Overview of referral-program skill

What the referral-program skill does

The referral-program skill helps an AI agent design, diagnose, or improve a customer referral program, affiliate program, or hybrid partner model. It is built for practical growth planning, not just brainstorming: it pushes the conversation toward program type, incentive structure, referral loop design, tools, and launch decisions.

Best fit for Product Marketing and growth teams

This referral-program skill is most useful for Product Marketing, growth, lifecycle, and founder-led teams that need to turn existing customers or partners into a repeatable acquisition channel. It is especially relevant when you need to answer questions like:

• Should this be referral, affiliate, or both?
• What reward fits our pricing and LTV?
• Why is participation low?
• What tooling and tracking do we need before launch?

The real job-to-be-done

Most users do not need a definition of referral marketing. They need a workable program that matches their product, economics, and customer behavior. This skill is strongest when you want help turning a vague goal like “get more word-of-mouth” into a concrete structure: trigger, share flow, conversion path, reward, metrics, and rollout checklist.

What makes this skill different from a generic prompt

The main advantage over a normal prompt is its built-in decision path. The skill explicitly checks for:

• referral vs affiliate vs hybrid fit
• B2B vs B2C context
• LTV and CAC constraints
• current participation and existing incentives
• product shareability and natural word-of-mouth fit
• tooling and budget reality

It also includes supporting references on affiliate design and real program examples, which helps the output move beyond generic “offer a discount” advice.

When this skill is a strong choice

Use the referral-program skill when:

• you are launching a new refer-a-friend or affiliate program
• you need to redesign incentives
• participation or conversion rates are weak
• you need examples, benchmarks, or tool suggestions
• you want a more structured referral-program guide for Product Marketing

When it is not the right tool

This skill is a weaker fit if your main problem is broad launch virality rather than an ongoing referral system. The repository itself points launch-specific virality questions toward launch-strategy. It is also less useful if you cannot share basic business inputs like pricing, LTV, customer type, and current acquisition economics.

How to Use referral-program skill

Install referral-program in your skills setup

The repository does not show an install command inside SKILL.md, so use the standard skills installer pattern from the repo root:

npx skills add https://github.com/coreyhaines31/marketingskills --skill referral-program

If your environment already supports local or synced skills, add it from skills/referral-program within the coreyhaines31/marketingskills repository.

Read these files first before first use

For a fast, high-signal read, start here:

1. skills/referral-program/SKILL.md
2. skills/referral-program/references/program-examples.md
3. skills/referral-program/references/affiliate-programs.md
4. skills/referral-program/evals/evals.json

That reading order matters. SKILL.md gives the workflow, the references add specifics, and evals/evals.json shows what a good output is expected to include.

Check for product marketing context before prompting

This referral-program skill explicitly tells the agent to look for .agents/product-marketing-context.md or .claude/product-marketing-context.md first. If you already maintain product positioning, audience, pricing, or GTM context there, the skill can ask fewer redundant questions and produce a more grounded plan.

If that file does not exist, give the same context directly in your prompt.

Give the minimum inputs that unlock a useful output

At minimum, provide:

• program type you think you need: referral, affiliate, or unsure
• B2B or B2C
• product and pricing model
• average customer LTV
• current CAC from other channels
• whether a program already exists
• current referral participation or conversion rates
• incentives already tried
• shareability of the product
• tools you use or can afford

Without these, the referral-program usage will stay generic because the reward economics and program mechanics depend heavily on them.

Turn a rough request into a strong referral-program prompt

Weak prompt:

Help me make a referral program.

Better prompt:

Use the referral-program skill to design a customer referral program for our B2B SaaS. We charge $49/month, average 14-month retention, CAC is about $180 on paid channels, and we have 1,000 active customers. We do not have an existing referral program. Customers often invite teammates, but not external peers. Recommend whether referral, affiliate, or a hybrid model fits best, propose incentives, define the referral loop, suggest tools, and include a launch checklist.

This works better because it gives the skill enough information to size incentives and judge fit.

Ask the skill to separate referral and affiliate logic

One practical strength of this repository is that it does not blur customer referrals and affiliate partnerships. If your business could support both, say so explicitly and ask for separate recommendations.

Example:

We have happy customers and industry creators. Use the referral-program skill to compare a customer referral program versus an affiliate program, including incentive structure, tracking requirements, fraud risk, and which should be launched first.

That usually produces a better decision than asking for one blended program.

Use the built-in workflow for program design

A solid referral-program workflow with this skill is:

1. establish context and economics
2. classify the program type
3. test product fit for sharing behavior
4. design the referral loop
5. choose incentives
6. select tools and tracking
7. define launch checklist
8. set success metrics and review cadence

If you skip steps 1 to 3, the rest often becomes unrealistic.

Use the references to pressure-test the output

The support files are worth using, not just browsing:

• references/program-examples.md helps compare incentive types, viral mechanics, and program styles
• references/affiliate-programs.md helps with commissions, cookie windows, affiliate recruitment, and fraud prevention

This is especially useful if the first draft feels too generic or too biased toward one model.

What good referral-program output should include

Based on the evals, a strong result should usually contain:

• a distinction between referral and affiliate models
• a proposed referral loop: trigger, share, conversion, reward
• incentive recommendations sized to your business model
• launch or rollout checklist
• tool suggestions
• optimization advice if you already have a program

If the output lacks those pieces, ask the agent to revise rather than accepting a shallow plan.

Practical prompt patterns for common use cases

For a new program:

Use the referral-program skill to design a first referral program for our product. Include eligibility, incentive structure, share flow, tracking, abuse prevention, and launch checklist.

For low participation:

Use the referral-program skill to diagnose why only 5% of customers have ever referred someone. Prioritize participation bottlenecks, messaging issues, reward mismatch, trigger timing, and product-fit constraints.

For affiliate setup:

Use the referral-program skill to propose an affiliate program for our subscription product, including commission model, cookie duration, recruitment channels, enablement assets, and fraud controls.

referral-program skill FAQ

Is referral-program mainly for referral programs or affiliate programs?

Both. The skill clearly covers customer referral programs, affiliate programs, ambassador-style models, and hybrid cases. Its value is partly in helping you choose the right one instead of assuming all partner-led growth works the same way.

Is the referral-program skill good for beginners?

Yes, if you can provide business basics. You do not need to be an expert in growth loops, but you do need enough context for the model to reason about incentives and channel economics. Beginners get the most value when they ask for a recommendation plus explanation of tradeoffs.

What does this skill do better than a normal AI prompt?

A normal prompt may produce generic incentive ideas. The referral-program skill gives the agent a more structured checklist and reference base, so it is more likely to ask the right setup questions, distinguish referral from affiliate mechanics, and include launch details rather than stopping at strategy slogans.

Is referral-program useful for Product Marketing specifically?

Yes. referral-program for Product Marketing is a good fit because Product Marketing often owns positioning, audience understanding, messaging, and lifecycle triggers that affect referral adoption. This skill is most useful when PMM needs to connect product value, customer motivation, and acquisition economics.

Does it help with optimization, not just new launches?

Yes. The evals show a clear optimization use case around low referral participation. That makes this more than a one-time planning document; it can also help diagnose weak adoption, poor incentive fit, or friction in the referral flow.

When should I not use referral-program?

Do not rely on this skill if:

• your product has little natural sharing behavior and you have no reason to believe incentives can overcome that
• you need highly technical implementation details for a specific referral platform
• your problem is broader launch virality rather than an ongoing referral-program system
• you cannot share even rough economics, pricing, or customer context

Does the skill include benchmarks or examples?

It includes curated examples and design references rather than a large benchmark database. The examples are useful for pattern matching incentive types, program structure, and viral mechanics, but you should still adapt them to your own pricing and audience.

How to Improve referral-program skill

Start with economics, not reward ideas

The most common mistake in referral-program usage is jumping straight to incentives. Better outputs start with LTV, CAC, retention, margin, and customer type. That is what lets the skill suggest whether a free month, account credit, flat payout, or recurring commission is financially sensible.

Be explicit about natural sharing behavior

The skill asks whether the product is shareable, has network effects, or is naturally talked about. Answer this honestly. If customers rarely discuss the product, the right result may be a modest referral test or an affiliate program, not a large incentive rollout.

Describe the customer moment when a referral ask could happen

Referral programs live or die on timing. Include the trigger moment:

• after activation
• after a success milestone
• after NPS or positive feedback
• after team invite or collaboration event
• after a repeat purchase

This helps the skill design a believable referral loop instead of a generic “add a share button” plan.

Give actual constraints on tools and ops

If you need a lightweight tool, say so. If finance can only support monthly payouts, say that too. If legal needs approval for rewards, mention it. The better your constraints, the more useful the referral-program guide becomes.

Ask for output in a decision-ready format

You will usually get better results by requesting sections like:

• recommended model
• incentive options with pros and cons
• referral loop
• tooling
• launch checklist
• success metrics
• risks and fraud controls

That format makes the output easier to review internally and easier to compare against alternatives.

Use the examples to force specificity

If the first answer is abstract, ask the agent to revise using one or two repository examples as anchors:

Rework this referral-program recommendation using Dropbox-style double-sided value and Morning Brew-style gamified milestones, but adapted to our SaaS economics.

This tends to produce sharper incentive and flow design than a generic retry.

Watch for these common failure modes

Weak referral-program outputs often:

• confuse customer referral with affiliate recruitment
• recommend rewards without checking unit economics
• ignore participation friction
• assume all products are naturally shareable
• skip the referred-user experience after the click
• omit fraud or abuse controls

If you see any of these, ask for a revision targeted at that gap.

Improve low-quality first drafts with sharper follow-ups

Good follow-up prompts include:

• “Revise this for B2B with longer sales cycles.”
• “Make the incentive safer for a low-margin subscription.”
• “Show 3 reward options ranked by ROI risk.”
• “Add fraud prevention and tracking requirements.”
• “Diagnose why participation is low before suggesting bigger rewards.”

These improve output quality because they force tradeoff thinking, not just more ideas.

Validate against the eval expectations

A practical way to improve referral-program results is to compare the output against evals/evals.json. If your answer does not check for product-marketing context, separate referral from affiliate logic, define the loop, size incentives, and include launch guidance, it is probably below the repository's intended quality bar.

Treat referral-program as a structured advisor, not an autopilot

This skill is most valuable when you use it to accelerate decision-making, not outsource judgment. The best results come from giving strong business context, reviewing the proposed mechanics critically, and iterating once the first program draft exposes the real tradeoffs.