Compare groups in results
Compare group in Results Studio focuses analysis on one experiment arm or variant assignment created by logic such as randomization groups or compare assets.
When you will see it
Section titled “When you will see it”- Randomization groups assign pages or blocks.
- Compare assets shows different prototypes or creatives.
- Experiment recipes combine branching and randomization.
If the study has no assignments, the control may be hidden or show a single default group.
How to use Compare group
Section titled “How to use Compare group”- Open Results for the study.
- Locate the Compare group selector (often near filters on Dashboards or Summary).
- Pick an arm (for example Variant A).
- Review metrics, charts, and sessions for that arm.
- Switch arms and note differences before you write recommendations.
- Optionally combine with filters such as device or completion.
What to compare
Section titled “What to compare”| Signal | Why |
|---|---|
| Completion rate | One variant may cause drop-off |
| Task success / time | Usability differences |
| Preference / rating means | Directional product choice |
| Open text themes | Qualitative explanation |
| Recording samples | Behavioral proof |
Example
Section titled “Example”A homepage hero compare-assets test:
- Select Variant A; note bounce-like early exits on Summary.
- Select Variant B; confirm higher completion.
- On Responses, open three recordings per arm.
- Export each arm’s Current view for the design review.
Limitations
Section titled “Limitations”- Unequal sample sizes across arms reduce confidence. Check counts first.
- Changing logic mid-flight can mix old and new assignments; Pause when redesigning experiments.
- Compare group is analysis-time slicing; it does not rewrite stored answers.
Related
Section titled “Related”Common questions
Where does Compare group come from?
It appears when the study uses randomization groups, compare assets, or related experiment setups that assign participants to arms.
Can I see all groups at once?
Use the compare selector and charts that support breakdowns. Start with one arm when validating data quality, then compare.
Do synthetic responses respect groups?
When synthetic data is generated for experiments, use the Real or Synthetic toggle carefully and confirm arm labels before presenting.