Referral and Affiliate Programs Playbook

A referral program turns happy customers into a channel; an affiliate program pays outsiders who already have your buyers’ attention. Both work when the product gives people something worth mentioning, the reward costs less than any other way you buy a customer, and every referral is tracked to a payment. Ask at the moment of value, pay a modest two-sided reward after the first payment, disclose every paid connection, and count the loop stage by stage.

Founders ask me for “a Dropbox-style referral program” as if the reward were the engine. In my view it isn’t: a reward speeds up word of mouth that is already happening, and can’t create it. So the first job is to find out whether people already recommend you.

When it works, the customers it brings tend to be worth more. Tracking about 10,000 customers of a leading German bank for almost three years, a 2011 study found the average referred customer was worth at least 16% more than a non-referred one, with a retention advantage that persisted^[1]. And referred customers refer more: across field data sets in finance, software and retail, a 2024 study found they made between 31% and 57% more referrals than customers who joined another way^[2].

This playbook is the system around the program: the fit check, the ask, the reward, the tracking, the affiliates, the rules and the review. For depth, read Referral Programs for Startups and Affiliate Programs for B2B SaaS.

Before you build anything, look for evidence of word of mouth. Add “How did you hear about us?” to sign-up as an open text box and count the answers that name a person; the Self-Reported Attribution Coder sorts them. If almost nobody names a colleague, fix the product or onboarding first.

Then decide who will do the recommending (Figure 1). A customer referral program fits when customers already name colleagues and a referred person can start without a salesperson. An affiliate program fits when consultants, newsletter writers or educators already reach your buyers. When the referrer is a company that sells or implements beside you, the Partner Co-Marketing Playbook is the better guide.

Start with one program. In my view, a seed company that launches both on the same day runs neither well.

A link in the account settings is a program in name only, because nobody goes looking for it. Tie the ask to an event in the product: the first time it delivers its main result, a milestone, a renewal or an upgrade. Show it on the screen right after the moment, and follow up once by email a day later (Figure 2).

Ask customers who have reached your activation moment and are still active; the Customer Onboarding and Retention Playbook shows how to define it. Don’t ask in the first week of a trial: the customer has nothing to vouch for yet.

Remind referred customers how they joined. In a field experiment with more than 10 million referred customers, a reminder that they had been referred raised their referrals by more than 20%^[2]. One line does it: “You joined because Sam invited you. Who else would get value from this?”

The reward is a cost of acquiring a customer, so price it like one. Three numbers set the limit: the most you can pay for a customer (gross profit a month × your payback target in months, which the Unit Economics Calculator works out), what your cheapest working channel pays today, and the share of referred customers who would have bought anyway. Divide the reward by the share who would not have, to get the cost per customer the program actually adds.

Bigger isn’t better. In a field experiment with more than 160,000 bank customers, larger referral rewards brought more new customers but “considerably decrease[d] the profitability of referred new customers”^[3]; the authors checked the finding against archival data on about 270,000 telecom customers.

Four design rules follow from that. Reward both sides, so the friend has a reason to start now. Prefer account credit or a free month to cash, because it costs you less than its face value and attracts people who want the product. Pay after the first payment, not the sign-up, and cap the reward per customer per year.

If a referral can’t be traced to a paying account, you can’t pay the reward fairly or tell whether the program works. Give each customer a personal link, a short code for when the link gets lost, and a prefilled message they send from their own email or LinkedIn. Let them send it themselves rather than collecting colleagues’ addresses for your system to email.

Store the referrer on the new account at sign-up, on your server, from the link or the code. A browser cookie alone isn’t enough: privacy features in browsers can clear it before a trial ends, as the affiliate article explains in its section on tracking.

Write the credit rules down before launch: what counts, who gets credit when two channels touch the same customer, and what doesn’t count (Figure 4). Paying on the first invoice, not the sign-up, removes the easiest fraud.

At seed, an affiliate program is a handful of people who already reach your buyers, not an open sign-up page. Start with the names in your “How did you hear about us?” answers, then consultants who implement tools like yours, newsletter writers your buyers read and educators who teach the job your product does. Approve each one by hand.

Set the commission from your own numbers, so the expected cost per paying customer, after churn and overhead, sits under your ceiling. Then compare it with published terms: HubSpot, for example, publishes a “30% monthly recurring commission for up to one year and a 180-day cookie window”^[4]. Published terms tell you the market, not your margins.

Write the terms on one page before the first link goes out (Figure 5), and give affiliates what makes them effective: a demo account, a brief on who the product is for, approved claims and one person to email. Collect tax forms before the first payout; the affiliate article covers the US forms.

A paid recommendation is still an advertisement. The FTC’s Endorsement Guides say that when a connection between an endorser and the seller “might materially affect the weight or credibility of the endorsement” and isn’t reasonably expected by the audience, “such connection must be disclosed clearly and conspicuously”^[5]. Two of the Guides’ examples fit these programs directly: points awarded for posting about a product, and a blogger paid a share of each sale through links.

Keep rewards and reviews apart. The FTC’s rule on consumer reviews, in force since October 21, 2024, bars buying positive or negative consumer reviews^[6], so never make a reward depend on a rating. The UK and EU have their own rules on labelling paid content, and selling to hospitals, clinics or public schools brings gift and anti-kickback rules; the two AAJ articles cover both. This is orientation, not legal advice.

Before each payout, check new referred accounts for self-referrals (the same card, email domain or device as the referrer), search ads on your brand name with an affiliate link, codes posted on discount sites, and accounts that cancel right after the reward. Void the reward or remove the affiliate under the terms.

A referral program is a loop, and its weakest stage limits all of it. Count five stages each month: customers asked, customers who shared, people invited where you can count them, trials started, and paying customers who passed the credit rules (Figure 7).

Add three money numbers: rewards paid, cost per added customer, and the retention of referred customers against everyone else. If referred customers churn faster, the reward is attracting the wrong people; the Churn and Retention Calculator makes the comparison. For affiliates, check each one monthly, and drop anyone who sends no customer in a quarter.

Read the loop honestly. In my view, B2B referral is usually a cheap addition to other channels, not a self-sustaining engine, so divide new referred customers by active customers before deciding how much more effort it deserves.

What changes by stage?

What mistakes should I avoid?

• Launching a reward before anyone recommends you. A reward can’t fix a product people don’t mention (Step 1).
• Hiding the program in settings. Nobody goes looking for it (Step 2).
• Rewards bigger than your cheapest channel. You buy the same customers at a higher price (Step 3).
• Paying on sign-up. You pay for accounts that never become customers (Step 4).
• An open affiliate sign-up page. Coupon sites and brand bidders arrive before the people your buyers trust (Step 5).
• Paid recommendations without a disclosure. The connection has to be clear, near the link (Step 6).
• Tying a reward to a review. Buying positive reviews is banned in the US (Step 6).
• Counting links shared. Judge the program on paying customers and their retention (Step 7).