How do I design a referral program that works?

A good referral program gives both people something they value, pays out only after the new person gets real value, and makes sharing and tracking take seconds.

So the design has four parts:

1. The reward: what each side gets, in a unit your product already uses.
2. The trigger: the event that earns the reward.
3. The guardrails: rules that stop people farming rewards with fake accounts.
4. The screens: a share flow and a progress tracker.

The best evidence comes from a study in the Journal of Marketing. Schmitt, Skiera, and Van den Bulte (2011) tracked about 10,000 customers of a leading German bank for almost three years. Referred customers:

• had a higher contribution margin, though the gap eroded over time;
• had a higher retention rate, and that gap persisted;
• were worth at least 16% more on average than similar non-referred customers.

The authors add a warning: the gap varied across customer segments, so firms "should use a selective approach." One bank is not your app, so read it as a signal.

The best-known product example is Dropbox. In his April 2010 Startup Lessons Learned slides, Drew Houston wrote that a "referral program w/ 2-sided incentive permanently increased signups by 60%," and that users sent 2.8 million direct referral invites in the trailing 30 days. That is the company's own report, not an independent audit.

At ImagineComputer (formerly Chatly) I rebuilt the referral system end to end. The offer was 50% off for referred users and 7,000 credits per upgrade.

Two choices in that design are worth copying:

• The new person gets something too. A discount makes the invite read as a gift, not a favour.
• The credit reward is tied to an upgrade, not a signup. And credits are the product's own currency, so the reward is something people use inside the product.

The results of that work are not public, so I will not quote any.

Research gives a clear steer here.

• Rewards raise referrals. Ryu and Feick (Journal of Marketing, 2007) ran four experiments and found rewards increase referral likelihood, especially to weak ties and for weaker brands.
• Who gets it depends on the relationship. In the same study, for weak ties and weaker brands, rewarding the person who recommends matters most. For strong ties and stronger brands, giving at least some of the reward to the receiver seemed more effective.
• One-sided rewards can look suspicious. Verlegh, Ryu, Tuk, and Feick (2013) found that rewards can make receivers suspect ulterior motives, more so for unsolicited and weak-tie referrals. Rewarding both sides, or offering symbolic rewards, "can eliminate the negative effect."

Most products are a weak brand to a stranger, and most invites go beyond close friends. Two-sided is the safe default.

Pay out after the new person has done the thing that makes them valuable, not when they click a link.

Dropbox is a good model. Its help centre says both people get the extra storage only after the friend creates an account through the link, installs the desktop or mobile app, logs in to the desktop app, and verifies their email.

Pick the earliest trigger that a fake account cannot easily hit.

If a reward is worth something, someone will try to farm it. Dropbox's published rules show the basic guardrails:

• Caps. Basic accounts earn 500 MB per referral, up to 16 GB. Plus, Family, and Professional accounts earn 1 GB, up to 32 GB.
• No existing users. A referral is ineligible if the friend already has an account.
• No self-referral from one machine. It is also ineligible if the friend registers from the same computer as the sender.
• One winner per friend. Only the sender of the link actually used earns the bonus.

On top of those, consider holding paid rewards until any refund window has passed, and watch your analytics for one account referring many accounts that never come back. Write the rules in plain words on the referral screen, so honest users know them too.

The share flow:

• Lead with what the friend gets, then what you get. "Give 50% off, get credits" is the order.
• One tap to copy the link, plus the share sheet on mobile. Do not make people write the message.
• Prefill a short, honest message the person can edit.
• Show it after a win, when the person has just got value, the same timing logic as a paywall. My paywall guide covers it.

The progress tracker:

• Show every invite and its status. Dropbox labels them with statuses like "Waiting for install," "Waiting for email verification," and "Completed."
• Show the total earned and the cap, so people know where they stand.
• Explain why a referral did not count instead of letting it disappear.

1. Pick your reward unit for each side: credits, storage, a discount, or time.
2. Choose the trigger using the table above, and write down why a fake account cannot hit it.
3. Sketch two screens: the share flow and the progress tracker, with every status.
4. Decide your metric before launch, and segment it so one abusive account cannot hide in the average.