Skip to content

Experiment recipes

Copy-paste starting points for common research designs. Adjust page counts and group names to match your study.


Goal: Everyone sees the same flow.

  • No randomisation on sections and pages
  • Optional display logic for screener skip or answer-based follow-ups

Goal: Reduce order bias without hiding content.

  • Randomise order at page level (blocks) or section level (pages)
  • Keep instructions on a separate page with no randomisation

Goal: Each person sees some modules, not all; group names not needed in results.

  • Split modules into a section
  • Present a subset → e.g. 2 of 5 subsections
  • Do not enable Randomisation groups

Classic A/B test (one variant per participant)

Section titled “Classic A/B test (one variant per participant)”

Goal: Group A vs Group B with named conditions in results.

  1. Create a section with one page (or subsection) per variant
  2. Present a subset → 1 of N
  3. Enable Randomisation groups → map Group A/B to each target
  4. Add shared post-task questions outside the randomised section
  5. Keep thank you page outside the section

Goal: Three or more between-subjects arms with balanced assignment.

  • Same as A/B with N pages and N groups (or N blocks on one page)
  • Use weighted randomisation if groups are not equal size
  • Test every arm in preview with Follow logic

Goal: Same task, different prototypes per group.

  1. Compare assets on task pages with randomisation driver
  2. Optionally combine with section randomisation if each group has a different task flow
  3. Define groups in the creation wizard or inline in compare assets

Goal: Condition comes from the URL, not random assignment.

  1. Create a URL parameter variable (e.g. ?condition=B)
  2. Display logic on each page or section: show when variable equals the right value
  3. Do not use section subset mode — assignment comes from the link

Mixed: random group + conditional follow-up

Section titled “Mixed: random group + conditional follow-up”

Goal: Between-subjects main task, then answer-driven paths.

  1. Randomisation groups for the main between-subjects block
  2. Shared later pages with display logic based on answers
  3. Branching to jump qualified participants to the main block