A/B/n split testing

Find the winning variant inside your journey

Split journey traffic across up to 4 branches, each with its own copy, timing, or channel. Pushwoosh scores every branch against your conversion goals and sends new users down the one that performs best.

Customer Journey canvas: a cart-abandoned trigger feeding an A/B/n Split element that sends 25% of traffic down each of four branches — the current push reminder, a new push copy, the same offer by email, and an equal delay with no message — all ending at a shared exit

Everything you need to run a split test

Up to 4 branches

Branch A is the control. You set the traffic share for each branch.

Random allocation

Every new entrant is assigned by the percentages you configure.

Goal-driven results

Track 1 conversion event or several, each scored separately.

Significance built in

Z-Score and P-Value, marked Significant at 90% confidence.

Automatic winner routing

New users are routed to the winning branch without a manual step.

In-flight users protected

Anyone already in a branch finishes the path they started.

What each branch can change

Drop the A/B/n Split element anywhere on the journey canvas and give it 2 to 4 branches. Each branch runs its own sub-flow, so a variant can differ by message copy, by delay length, by channel, or by sending nothing at all. Traffic percentages have to add up to 100%, and a reset control puts them back to an even distribution.

  • Branch A – the control, either the flow you run today or a fresh baseline.
  • Branches B, C and D – the challengers, each with its own messages, delays and channels.
  • Traffic share – any split that totals 100%. An even split reaches significance fastest.

The split element carries no messaging logic of its own. It decides who goes where, and the channel blocks underneath it do the sending.

Two ways to close a test

Automatic winner selection

The platform watches your goal continuously. Once a branch clears the 90% confidence bar, Pushwoosh deactivates the underperforming branches and sends every new entrant to the winner.

That happens whether or not anyone is looking at the dashboard.

Manual control when you want it

Manual mode surfaces the same live statistics and leaves the call to you. Read the numbers, then click Keep only winner to deactivate the losing branches.

The journey keeps running while you decide. Closing a test does not require pausing it.

Deactivation applies to new entrants only. Users already assigned to a branch complete the path they were given, whatever happens to that branch afterwards.

What A/B/n split testing is best for

Cart abandonment

Test reminder copy and delay length, then keep the version that recovers more carts.

Paywall and trial messages

A/B the wording or the timing of a trial-ending message against your conversion event.

Digest headlines and send times

Run 2 headlines or 2 send times on a recurring digest and score them on opens or clicks.

Channel preference

Send the same message on push, email and SMS to find out which one converts for a segment.

Monetization offers in games

Test offer copy or timing inside a re-engagement journey, with purchase events as the goal.

Win-back timing

Compare how long to wait before reaching out to a dormant user.

One journey, every channel

The split element sits on the same canvas as every channel block, so a branch can carry mobile push, in-app messages, email, SMS, WhatsApp or web push against a single customer profile.

You build the test where the journey already runs, inside the Customer Journey Builder, instead of wiring up a separate experimentation tool alongside it.

Three ways to design the test

Current flow vs a new idea

Branch A keeps what you send today. B, C and D carry the challengers.

New idea vs new idea

No incumbent involved. Pick one new version as the control and score the rest against it.

Message vs no message

Give one branch an equal Time Delay and no message, to measure whether the touch moves the metric at all.

How a winner is decided

Define at least 1 conversion goal before the test can launch. A goal is a tracked event you already send to Pushwoosh: a purchase, a checkout, a subscription start, a level completion.

Pushwoosh scores each branch against the control with a Z-Score and a P-Value, and marks the result Significant at a 90% confidence level.

Several goals can run on the same test at once. Results are calculated separately per goal, and you choose which goal to read when more than one is live.

Testing on infrastructure you can name

Pushwoosh runs on its own hardware in the US and Germany, under GDPR and BDSG. Your test data, goal events and customer profiles stay on the same platform that sends the messages.

ISO 27001:2022 CertifiedISO 27001 CertifiedGDPR CompliantData Privacy FrameworkHIPAA CompliantSOC 2 Type I CertifiedOWASP Compliant

Full certification detail on the data safety page.

How it works

  1. Pause the journey and add the split

    Place the A/B/n Split element on the canvas, set 2 to 4 branches, and give each one its traffic share.

  2. Define the conversion goal

    Pick the event that counts as a win. Check that it can fire in every branch, including the control.

  3. Launch and pick your mode

    Automatic routes new users to the winner on its own. Manual leaves the Keep only winner call to you.

Good to know before you launch.

  • The journey has to be paused to add or reconfigure a split. It is not a hot edit on a running journey.
  • Up to 4 branches per split element.
  • A goal that can only fire in one branch breaks the comparison. Every branch, control included, has to be able to complete it.
  • Significance is called at 90% confidence. That bar names a winner sooner than a 95% one and carries a higher false-positive risk.
  • Pushwoosh does not enforce a minimum sample size or a minimum test duration before a result can be marked Significant. Low-traffic journeys need patience.