Overview
Loyalty programs are harder to test than referral campaigns, because the outcome you care about happens weeks after the change. A member has to earn, accumulate, and redeem before a test tells you anything. This page covers what is worth testing, and how to run a test when the feedback loop is long. For the mechanics of setting a test up, see A/B Testing Your Campaign and A/B Testing Your Offer.What to Test
Earn Structure
Campaign variants can isolate settings owned by the campaign:- Earn rate. Link each variant to a different Collectible Reward value and compare participation and cost.
- Reward behavior. Compare First Action with Every Action for a repeatable event.
- Eligibility rules. Compare which audience or event-data conditions qualify for points.
Ways to Earn
- Which actions to reward. Adding a way to earn should increase the action. If participation does not move, you are paying for behavior you were getting anyway.
- Which actions to feature. Change which way to earn the creative promotes while leaving the underlying Reward Bank configuration unchanged.
Messaging
The fastest-reading tests, and the ones to run while an earn-structure test matures.- Earned reward emails. Whether leading with the balance or with the reward drives more visits to the Redemption Center.
- Reminders. Whether prompting members with an unspent balance increases redemption.
- Enrollment. How the program is described at sign-up, and what that does to join rate.
Reward Bank Settings
The Redemption Ratio, the minimum and maximum redemption values, Redemption Options, and the Redemption Center belong to the Reward Bank extension. Duplicating a campaign does not create a second bank or isolate those settings, and both variants continue to add value to the same member balance. Testing bank-level settings requires separate Reward Banks with distinct Zone Name values and explicit targeting so each member reaches the correct bank. That also separates balances, so plan the migration and member experience with your Extole team before running this kind of test.Running a Test
Set the test up by duplicating the campaign, changing one thing in the variant, and taking it live as a test. Traffic splits 50/50, and existing participants continue in the campaign they are already in. Compare the campaigns side by side on the Program Dashboard.Reading a Result
Pick the measure before you start, and pick one that can actually move in the time you are prepared to wait.
Do not read an earn-structure test on an early measure alone. A variant that lifts enrollment and
then issues points nobody redeems has not won.