How do you build a referral program that works?

Reward both sides

A one-sided reward asks your user to spend social capital so that you and they benefit and their friend gets nothing. That framing is uncomfortable, and it shows in conversion.

Two-sided changes the message from I get something if you sign up to here is something for both of us. Dropbox gave storage to referrer and referred alike; PayPal paid both. In both cases the invitation became a gift rather than a solicitation.

Make the reward native to the product

Cash and gift cards work, in the narrow sense that they produce signups. They also select for people whose motivation is the cash, and those users churn quickly and refer more people like themselves. The programme grows a population that is expensive and worthless at the same time.

Product-native rewards — more storage, an extra seat, an unlocked capability, extended access — filter for people who want the product. They are usually cheaper for you, and they make the referred user's first experience better, which raises the chance they refer someone in turn.

Time the ask to realised value

The single most common design error is asking at signup. The user has not yet decided whether they like the thing you are asking them to recommend.

Place the ask immediately after a success: a project completed, a file shared, a first payment received, a milestone reached. Calendly's growth works partly because the ask is structural rather than prompted — every meeting link sent is an exposure at exactly the moment the tool has just been useful.

Reward on activation, not on signup

Whatever action triggers the payout will be optimised by people who want the payout. Pay on account creation and you will get farmed accounts. Pay on a meaningful action — activation, first transaction, thirty days retained — and the incentive points at the outcome you actually want.

This also protects the referrer relationship. Nobody wants to be told their friend signed up but the reward is withheld; state the condition plainly up front.

Measure the right things

• Share of new users from referral, not the raw referral count.
• Retention of referred users versus other channels. Referred users often retain better; if yours retain worse, the incentive is attracting the wrong people.
• Cost per activated referred user, compared against your other channels. A referral programme is an acquisition channel and should be judged against the alternatives.
• Participation rate. If 2% of users ever refer anyone, the problem is the ask or the timing, not the reward size.

When not to build one

If your retention curve does not flatten, a referral programme accelerates the leak. Referral amplifies whatever the product already does — it makes a good product spread and a mediocre one reach its ceiling faster while burning goodwill.

Fix retention first. A programme launched on a product people quietly abandon converts the most valuable thing you have, your users' willingness to vouch for you, into nothing.