A/B Testing in Journey Builder for Salesforce Marketing Cloud

Testing a subject line or a send time is easy. Optimizing an entire journey — from the welcome email through a conditional follow-up three weeks later — is a different game. In Salesforce Marketing Cloud, A/B testing isn't confined to Email Studio: Journey Builder lets you compare variants of content, timing, and even logic paths inside a live automation. This guide shows you how to build a reliable A/B test with the Random Split and Decision Split activities, how to call a winner from the Data Views, and what the Summer '26 release changes about the way you manage your journeys.

Why test in Journey Builder instead of Email Studio

Email Studio's native A/B test is great for a one-shot send: two subject lines, a winning criterion, and automatic delivery of the champion. But it stops at the email. The moment your question becomes "which sequence keeps subscribers engaged across three weeks?", you have to think at the journey level.

Journey Builder gives you that resolution. You can test two full sequences — cadence, channels, content — and measure the impact on a downstream conversion rather than an immediate open rate. That's the difference between optimizing a message and optimizing an experience.

Random Split: building a clean A/B test

The Random Split activity randomly distributes incoming contacts across multiple paths. It's the foundation of any journey-level A/B test. Configure two 50/50 outcomes (or 90/10 if you want to cap the exposure of a riskier variant), then place a different sequence on each branch.

Sizing the sample correctly

A test is only worth running if it's statistically readable. Before you launch, estimate the volume you need to detect the effect you're chasing. A practical rule: to spot a 2-point lift on a 10% conversion rate, plan for several thousand contacts per branch. Below that, your "winner" is usually just noise.

Also lock your population: don't inject new contacts mid-test, or you'll be comparing different cohorts and your result will drift.

Decision Split: routing on real behavior

Where Random Split distributes at random, Decision Split routes on attributes or events. After sending your two variants, use a Decision Split to separate the contacts who clicked from those who stayed silent, and tailor the rest of the journey accordingly.

A Decision Split criterion draws on contact or engagement data. For example, targeting contacts who opened the email within the last 48 hours:

/* Conceptual Decision Split criterion on engagement */
Contact.Attribute.Engagement.Opened = True
AND Contact.Attribute.Engagement.OpenDate >= AddDays(Now(), -2)

Upstream, you can precompute a segmentation flag in a Data Extension with a SQL Query Activity, then consume it in the Decision Split:

SELECT
    s.SubscriberKey,
    CASE WHEN o.EventDate IS NOT NULL THEN 'Engaged' ELSE 'Dormant' END AS EngagementTier
FROM   _Subscribers s
LEFT JOIN _Open o
       ON o.SubscriberKey = s.SubscriberKey
      AND o.EventDate >= DATEADD(day, -7, GETDATE())

Measuring and calling a winner

Journey Builder surfaces journey metrics, but for a rigorous verdict, lean on the Data Views (_Sent, _Open, _Click, _Job). Tie each send back to its branch so you compare variants on the same denominator.

SELECT
    j.EmailName,
    COUNT(DISTINCT s.SubscriberKey)            AS Sends,
    COUNT(DISTINCT o.SubscriberKey)            AS Opens,
    COUNT(DISTINCT c.SubscriberKey)            AS Clicks,
    CAST(COUNT(DISTINCT c.SubscriberKey) AS FLOAT)
      / NULLIF(COUNT(DISTINCT s.SubscriberKey),0) AS ClickRate
FROM   _Sent s
JOIN   _Job  j ON j.JobID = s.JobID
LEFT JOIN _Open  o ON o.SubscriberKey = s.SubscriberKey AND o.JobID = s.JobID
LEFT JOIN _Click c ON c.SubscriberKey = s.SubscriberKey AND c.JobID = s.JobID
GROUP BY j.EmailName

Don't stop at click rate: always bring the analysis back to the final business conversion — a purchase, a booked meeting, a completed form — because that's the only true measure of a journey's success.

An A/B test in Journey Builder is only as good as its question: change one variable at a time, judge it on real conversion, and let it run long enough that chance stops making the call for you.

What Summer '26 changes for your journeys

The Summer '26 release strengthens the Journey Dashboard: multi-select to move several journeys into a folder in a single action, and a new campaign filter to instantly surface the journeys tied to a given initiative. In practice, this makes it easier to organize your test variants — group "Variant A" and "Variant B" inside a campaign folder and track them without getting lost in a flat list.

These are productivity gains, not methodology: statistical rigor is still on you. But good organization cuts down on the handling mistakes that so often invalidate a test.

Best practices and mistakes to avoid

Never change two variables at once — you won't know which one produced the effect.
Document the hypothesis before launch, not after you've seen the numbers.
Wait for the conversion window to close before concluding — a late click can flip a premature verdict.
Re-run the winning variant on a fresh cohort before rolling it out everywhere.

Want to industrialize journey testing in Salesforce Marketing Cloud? Get in touch with CGC-Agency to frame your experimentation program.

Key takeaways

Test the journey, not just the email. Journey Builder measures how a full sequence affects a downstream conversion, where Email Studio's native A/B test stops at the first send.

Random Split for exposure, Decision Split for routing. The first distributes contacts at random across variants; the second steers them based on real subscriber behavior.

The verdict lives in the Data Views. Join _Sent, _Open and _Click per branch, and always trace the analysis back to the final business conversion.

Summer '26 makes organization easier. The campaign filter and multi-select on the Journey Dashboard help you tidy your variants — without replacing statistical rigor.