Skip to main content
Offer experiments compare live variants inside one recovery workflow. They do not replace continuous holdout measurement. Stripe variants pair an optional script direction with a percentage. RevenueCat variants pair it with an App Store Connect offer.

​
Create variants

Open Script & voice for an eligible workflow, then open its offer experiment controls.
  1. Turn on the experiment.
  2. Create two through five variants.
  3. Give each variant an offer and optional script direction.
  4. Save the workflow to start the experiment.
Saving starts the experiment. hyzl splits new eligible recovery traffic evenly across variants. A retry stays with the variant assigned on the first attempt.

​
Stripe variants

Set each variant’s percentage independently. A 0% variant creates a feedback-only conversation with no discount. A 0% variant cannot send a discounted resubscribe link because it has no discount to attach.

​
RevenueCat variants

Choose an App Store offer for each variant. Create missing discounts in App Store Connect through the provided setup route. Every variant needs an offer and usable codes before the experiment can start. Keep enough inventory available for all variants.

​
Protect the comparison

Offer and script inputs lock after the experiment starts. Close the experiment before changing them. Changing an active variant would mix old and new treatments in one result. Start a new experiment after any adjustment. Change one important element at a time when you want an interpretable result.

​
Read results

Open the relevant trigger in Analytics and find Offer performance. Results remain visible while an experiment runs and after it closes. Each variant reports available measures such as:
  • traffic allocation and calls
  • members saved or converted
  • save rate
  • monthly revenue retained
  • one-time recovery revenue collected
The highest monthly retained revenue can receive a Leading label when it is clearly ahead. A tie receives no label. Several experiments can appear in one time window. Analytics groups each experiment so their results never blend. When several triggers or providers contribute, tabs separate results. Labels can identify a trigger and provider. Older ungrouped data can appear under Earlier. Click View results in the workflow editor to open the matching Analytics result.
Offer performance reports raw counts, not statistical significance. Treat a small-sample difference as direction, not proof.

​
End an experiment

Choose the end action from Offer performance in Analytics. You can also close without a winner from the workflow editor.

​
Promote one variant

Click Promote winner and choose one variant. The experiment ends, and that offer becomes the workflow’s standing offer. The promoted script direction does not automatically rewrite the base script. Copy it into the script when you want to keep it.

​
Relaunch several variants

Choose two or more variants, but not all variants. hyzl closes the current experiment and starts a new one with the chosen variants. Traffic splits evenly across the new set. Keeping two of five gives each survivor half of new eligible traffic. The old result remains available. Relaunching creates a clean comparison instead of changing allocation inside a running experiment.

​
Close without a winner

Click Close test in Analytics or the workflow editor. Confirm the action. Recovery immediately returns to the workflow’s standing offer. Variants unlock so you can edit or duplicate them. Nothing is prechosen in the decision dialog. Review its outcome before confirming. Continuous lift compares recovery outreach with a standing holdout. It remains separate from offer experiments.