Skip to main content

Run an experiment

Updated October 10, 2026

The Experiment board shows A/B/C tests, such as two hooks or two posting times, with confidence intervals. A winner is only called when the result is statistically significant, so you don't act on chance.

Note: There is no button to create an experiment in the app yet. Experiments appear on the board once they are started, for example from the suggestions in the Weekly Analyst. To compare whole personas side by side, launch a cohort instead.

Read the board

  1. Open Analytics → Experiments.
  2. Each experiment shows its status (Draft, Running, Completed or Stopped), its Hypothesis, the variable it tests (for example Hook, Format, CTA or Posting time) and its Primary metric.
  3. The table lists each variant with impressions, views, conversions and the 95% CI (confidence interval). The chart shows each estimate and its interval.

Understand the result

  • "Variant … is the winner: statistically significant": you can act on it.
  • "No significant difference": the variants differ by no more than chance explains. Don't pick one based on this.
  • "Insufficient data": each variant needs a minimum number of published posts before a result is computed. No conclusion yet.

Results are recalculated once a day while an experiment runs. A significant result also appears in Intelligence → Recommendations as an experiment result.

Plan your next test

  • Use Next week's experiments in the Weekly Analyst as a starting point.
  • Change one thing at a time (hook, format, length, topic, CTA, offer, posting time or frequency).
  • Publish enough posts per variant; small samples rarely give a result.

What can block you

  • The analytics feature may not be enabled for your workspace.
  • "No experiments yet": nothing has been started for this workspace.

More in Analytics and decisions