Real split tests, not a hunch.
A controlled comparison — the same content, run at different times — with an honest per-arm result. Part of the AI Learning Engine's Insights tab, not a separate testing tool you have to set up.
Same content, two arms
Why "controlled" matters.
The Learning Engine distinguishes between what's observed, what's correlated, and what's actually been tested — and labels each result honestly.
A real experiment, not a hunch
Same content, deliberately posted at two different times or with two variants — a genuine controlled comparison, not just watching what happened to naturally perform well.
- Real per-arm engagement data
- Honest — no result until it's actually run
Labeled by evidence strength
A completed split test is tagged "Experimentally Supported" — a distinct evidence tier from "Observed" or "Correlated" claims elsewhere in your Insights tab, so you always know how much weight to put on a result.
- Never called "causal" without a real experiment behind it
Feeds the Learning Engine.
Results don't just sit in a report — they inform future recommendations for your account, the same evidence-backed system behind related features below.