A/B Testing

Stop guessing. Let your users decide.

Send two versions of a push, keep a control group, and let Lotse tell you which one makes people buy.

Why test in Lotse

Experiments on the data you already trust.

Measured on real purchases

The goal is an event your app already sends — not opens or clicks.

An honest control group

Part of the audience gets nothing, so you see the real effect of the push.

A winner you can trust

Lotse calculates how likely B beats A, and only then sends the winner to everyone.

How a test runs

One trigger. Two variants. One winner.

Decided by data

Know when it’s really decided.

Lotse waits for fixed minimum numbers before it names a winner. Then the winner goes out on its own or on your click.

What makes it different

Testing without a second tool.

No SDK update

Journeys use the events and push your app already sends, so they work for apps already in the store.

Audiences from anywhere

Target by origin campaign or behaviour — the same segments as everywhere in Lotse.

Each device once

A device enters a journey at most once and stays in its variant.

For developers

Nothing to ship.

Journeys run on the server. Your app only needs the events and push handling it already has.

Dart
// The goal of a journey is an event your app already sends.
Lotse.instance.track('purchase', props: {'value': '59.00'});

// push_open with the campaign comes from the SDK on its own.
Trigger
An install (a new device) or a named event, optionally with one property comparison.
Variants
One or two, plus an optional control group from 0 to 50 percent.
Winner
The goal rate per variant and the probability that B beats A, with fixed minimums.

Developer overview

Questions about A/B testing

What can I test?

The text, title and link of a push in a journey: two variants and an optional control group.

Can I test more than two variants?

Not in this version — A, B and a control group.

Can I test in-app messages?

Push first. In-app messages as an action come later.

A/B Testing

Find out which message really works.

Set up a journey with two variants and a control group.