Study groups

The most common confusion in group comparison studies is the difference between random sample (anonymous subset) and randomisation groups (named assignment). They can look the same to participants but behave very differently in results.

Random sample (default subset behaviour)

“Pick K items from this list at random.”

When you use Present a subset without enabling Randomisation groups:

Each participant randomly sees K pages, subsections, or blocks from the pool
There is no persistent group label (no “Group A” stored)
Assignment is local to that page or section
Fine when you only care that exposure was balanced, not which named arm someone was in
Not ideal when you need to filter results by “Variant A vs B vs C”

Randomisation groups (named assignment)

“Assign each participant to Group A / B / C and show the matching content.”

Enable Randomisation groups inside Present a subset mode (on a page or section). You can:

Select an existing randomisation variable, or
Create one inline (same flow as Compare assets) with group names and optional weights

Then map each group to a page, subsection, or content block.

What you get:

A stable group assignment per participant (stored on responses)
Server-side permuted-block balancing for live responses (fair splits even at small sample sizes)
Managed display rules on each target (read-only Randomisation group exposure in Display Logic)
Group values available in results, exports, and dashboards

Use when:

A/B or multivariate tests where results must be split by condition
Between-subjects prototype or task comparison
Any design where you export or compare by group name

Side-by-side example

Setup: Section with Page A, Page B, and Page C.

Intent	Setting
”Each person randomly sees 1 page; I don’t need group names in results”	Present a subset → 1 of 3 — without Randomisation groups
”Group A always sees Page A, Group B sees Page B, Group C sees Page C — and I need that in results”	Present a subset → 1 of 3 → enable Randomisation groups → map A→Page A, B→Page B, C→Page C
”Everyone sees all three pages but in random order”	Randomise order (not subset)

With 2 pages and “1 of 2”, random sample and randomisation groups can look the same to the participant — but only groups give you a named condition for analysis.

Weighted groups

When creating a randomisation variable inline, you can set weights for unequal group sizes (e.g. 50% / 25% / 25%). Live assignment uses server-side permuted-block balancing to honour weights fairly over the study.

Participant groups (usability test wizard)

When creating a usability test, the wizard Groups step can define participant groups used with:

Randomisation variables
Compare assets
Section or page display logic

This is study-wide setup — it works alongside Path Logic randomisation and display rules, not instead of them.

See Creating a usability test.

Randomisation groups vs display logic

Approach	Best for
Randomisation groups (in Randomiser)	Random A/B/C with automatic rule compilation and results tracking
Manual display logic	Custom conditions, multiple rules, URL/device/answer routing
Both	Random main experiment + conditional follow-ups based on answers

For URL-based assignment (?variant=B), use Display logic and routing — not subset mode.