Skip to Content
GuidesJourneysSplit (A/B) node

Split (A/B) node

The split node (type: experiment_split) distributes contacts randomly across 2-5 paths. Use it to test different messages, channels, wait durations, or entire sub-flows against each other.

How it works

  1. A contact arrives at the split node
  2. The engine randomly assigns them to one of the configured variants based on the distribution percentages
  3. The contact proceeds down their assigned path
  4. Analytics track performance per variant

Configuration

  • Variant count — 2 to 5 (default: 2)
  • Distribution — percentage per variant (must total 100%)
    • 2 variants: 50/50 by default
    • 3 variants: 33/33/34 by default
    • Adjust to any split (e.g., 90/10 for a small test group)
  • Winner metric (optional) — open rate, click rate, conversion rate, or revenue
  • Test duration (optional) — after this period, the winning variant gets 100% of new traffic

Common patterns

Message variant test

Split → Send A (email variant 1) → exit → Send B (email variant 2) → exit

Test different subject lines, copy, or CTAs.

Channel test

Split → Send via WhatsApp → wait → measure → Send via Email → wait → measure

Discover which channel drives more conversions for a specific audience.

Flow test

Split → Send + Wait 1 day + Follow-up → exit → Send + Wait 3 days + Follow-up → exit

Test whether a shorter or longer cadence performs better.

Winner declaration

If you set a winner metric and test duration:

  • After the test period, the variant with the best metric wins
  • All new contacts entering the split go to the winning path
  • Contacts already in other paths continue their original path
  • Analytics show a clear “Winner” badge

If you don’t set a winner metric, the split runs indefinitely at the configured percentages.

What’s next

Branch nodeExit node