Most referral programs fail in configuration rather than in concept. Here is the short version of what to set before you publish, based on the campaigns we see working.
1. Decide what the friend gets first
The reward to the new customer is the one doing the work. It has to be worth the awkwardness of recommending a shop to someone. A percentage discount on a first order is the reliable choice; a points balance is not, because a new customer has no relationship with your points yet.
2. Make the referrer’s reward conditional
Reward the referrer when the friend’s order completes, not when the link is shared. This is the default in the app and it is the default for a reason — the alternative pays out for activity rather than outcomes, and it is the first thing anyone gaming a program will find.
3. Set a floor, not a ceiling
A minimum order value on the friend’s discount protects your margin without capping anything. Set it somewhere near your average order value. Caps on the number of referrals, by contrast, mostly punish your best advocates.
Two defaults are worth leaving alone: reward on completed order, and the 30-day attribution window.
4. Turn on post-purchase, then stop
The order status page is the highest-converting placement in the app, because it asks at the one moment a customer has just decided to trust you. Enable it. Then resist adding the storefront popup on day one — run one placement, see what it does, and add the second once you have a baseline.
5. Write the share text yourself
The default copy is deliberately plain so it does not embarrass anyone. It is still generic. Two minutes rewriting it in your own voice measurably outperforms the default, and it is the only part of this list that requires judgement rather than a setting.
The two defaults most stores should leave alone
- Reward on completed order. Not on share, not on signup.
- The 30-day attribution window. Long enough to cover a considered purchase, short enough that you can read the results.
After launch
Leave it running for a fortnight before changing anything. Referral programs are slower to read than popups because they depend on a chain of two purchases, and most of the adjustments made in the first week are made on noise.