How to build a referral program people actually use

Why most referral programs get zero referrals

Here's the usual story. You add a "Refer a friend, get a month free" link, put it in the settings page or a welcome email, and wait. Nothing happens. A few months later you wonder whether referral programs work at all.

It's a common pattern. Another maker reached 1,000 registered users and not one of them signed up through a referral link. On a Product Hunt launch, a user asked how to invite friends, and the maker admitted the program existed but was "a bit hidden in our footer". When you read these threads side by side, the failures fall into four piles:

1. People are asked before they've gotten anything. An invite at signup asks someone to vouch for a product they haven't used yet. They won't.
2. The reward is for the wrong person or the wrong thing. Paying only the inviter makes the invite feel like a sales pitch. Paying for a signup pays for clicks, including fake ones.
3. The friend has no reason to accept. "My friend gets a month free" helps you. "You get a month free" helps the friend. Only one of those gets clicked.
4. It's buried. A link in the footer or the billing page is a link nobody sees at the moment they'd actually use it.

Notice what's missing from that list: the size of the reward. Most programs don't fail because $10 is too little. They fail because nobody asked the right person at the right time. And if people use your product once and never come back, fix that first, because a referral program can't spread something people don't keep using. Why users don't come back covers that part.

Referral, recommendation or affiliate: which one do you need?

These three get lumped together, and mixing them up is how people end up paying sales commissions to customers who'd have told a friend for free. They do different jobs:

One thread puts the affiliate problem well: at a $10 price point, a commission is too small to make a creator care. The better fit there is an in-product reward, like extra usage or credits for both people. On the other end, a founder whose single $79-a-month customer had brought in eleven more asked whether to start paying them. The replies said: thank them personally first. That's a recommendation from a fan, and turning it straight into a sales deal can change how it feels.

The rest of this post is about the first column: users inviting friends, with a reward for both.

How to create a referral program that works: five rules

1. 1

   Ask right after the first win

   The best moment to ask is the minute after someone gets a useful result: their first finished project, their first report, their first booking. That's when they're most sure it's good, and most likely to think of someone else who needs it. Put the invite on that success screen, not in settings.

2. 2

   Give both people more of what they came for

   Dropbox gave both people extra storage, the exact thing they signed up to get. Milk Road, a crypto newsletter, gave readers a relevant PDF for one referral. Credits in an AI app, an extra month of a plan, more projects or seats: rewards that make the product better are cheaper for you and worth more to the right people than cash.

3. 3

   Pay when the friend does their first useful thing

   Airbnb's $25 credit for each side was paid after the invited friend's first trip, not when they made an account. Do the same: the reward lands when the friend finishes their first project, sends their first invoice or makes their first order. As a bonus, fake signups never earn anything, so you don't need a fraud system on day one.

4. 4

   Cap it

   Set a limit per person, say 10 friends, and a monthly budget across everyone. Real users rarely hit it. The cap protects you from the one person who finds a way to farm it, and it keeps your worst-case cost a number you've already agreed to.

5. 5

   Make the invite one tap, with a reason for the friend

   One button that copies a link or opens a share sheet, with a short message written for the friend: what the product does and what they get. Airbnb found the email focused on what the friend would get did better worldwide than the one focused on the sender (no effect size was published, so take it as a direction, not a number).

What the famous referral programs actually show

Every referral post quotes the same three companies. Here's what the original sources say, and what they don't.

• Dropbox (2010). The deck reports a 60% permanent lift in signups, with referrals making up 35% of daily signups. Both sides got storage. These are company-reported numbers with no randomized test, and Dropbox was growing in other ways at the same time.
• Airbnb (2014). The redesign paid $25 to each side after the friend's first trip and made sending invites work on web, Android and iOS. The 300%+ figure compares the new program to Airbnb's own old one, not to having no program, and it doesn't show a company-wide revenue lift.
• Harry's (2013). About 85,000 valid emails in a week, roughly 77% through referrals, with prizes at 5, 10, 25 and 50 friends and a page that showed your progress. Twelve staff seeded their own networks after months of prep. It's a founder's account of email signups, with no data on how many bought.

The pattern across all three is the same: a benefit people actually wanted, for both sides, and an invite that took seconds. None of them proves that adding a referral page to your app will do anything on its own.

What is a referral worth? An honest cost worksheet

Before you pick a reward, check that it pays for itself. You need three numbers: the reward per side, the share of invited friends who reach their first win, and what a kept user is worth to you (monthly margin times how many months people usually stay).

Here's a worked example with sample numbers for a small SaaS. A kept user pays $12 a month for 10 months, which is about $96 after costs. The reward is 500 credits for each side, which costs you about $5 each in compute. Out of 100 friends who sign up through invite links, 40 finish a first project, and 10 of those become paying users who stay.

Same product, same reward, same friends. The only change is when you pay, and it's the difference between losing money and making it. Paying at signup also pays every fake account someone creates to collect the bonus. Paying at the first win doesn't.

Two honest caveats. Some of those ten friends might have found you anyway, so not all $960 is new. And measure what actually happens: count invited friends who reach the first win and are still around a month later, compared with everyone else who signed up that month. If referred friends leave faster, your reward is attracting the wrong people.

How it changes for a newsletter, a community and a shop

The five rules hold everywhere. What changes is what counts as the friend's first win and what reward fits.

A newsletter

A signup is cheap, so don't count it. Count a friend who confirms and opens a few issues. The reward works best when it's something only you could make: Milk Road gave a relevant PDF for one referral (the case is vendor-reported and doesn't separate referrals from its paid ads and partnerships). Recommendations from other newsletters are a different job: Workspaces got 67% of one month's new subscribers from them, per its platform's case study, with no data on who stayed.

A community

People invite friends to a server or a group that has a clear purpose and people who talk to each other. If new members join and only lurk, more invites won't help. The friend's first win is their first reply that gets a reply back. Reward both people then, with a role or perk rather than cash. How to grow a Discord server has the full playbook.

A shop

The first win is the friend's first order, after the return window. A discount for the friend and a credit for the sender is the classic double-sided setup. If you want people to keep inviting, add bigger prizes at 3 and 10 friends, the way Harry's did before launch.

Before you launch: a waitlist

With no product yet, the reward is getting in sooner. Robinhood's founder recalls 50,000 signups in the first week from a waitlist where inviting friends moved you up the queue, helped along by Reddit and Hacker News (founder-reported; the founder also says earlier products failed despite referral efforts). The promise did the work. The queue just made it easy to share. If you're at this stage, the first 100 users post comes before this one.

Set up a double-sided referral in Lurk

If you'd rather not build the ledger, caps and payouts yourself, this is one rule in Lurk. Your app sends an event when an invited friend finishes their first project, saying who invited them. The rule does the rest:

• Every credit lands with a receipt, so you can always answer "why did I get this?" The Referral template comes with views for top referrers and friends who stayed.
• Before it goes live, preview the rule on your last 30 days to see what it would have paid. Add a budget across everyone if you want a hard ceiling.
• Ask your coding agent to set it up over MCP. Agents can draft and preview, but only a person on your team takes it live. See Lurk for agents.
• Other shapes: a Guest pass gives members passes to hand out, a Waitlist jump moves people up the line, and Founding members gives your first 100 a numbered badge. For bigger prizes at more friends, there's the Referral ladder. More for builders on Lurk for apps.

Lurk won't make people want to invite their friends. That part is still your product. It just makes the right reward land at the right moment, every time, with a receipt.