Within vs. Between Design

When you build a study from scratch, the first choice is how your audience meets the things you’re testing.

• Single group — within-subject design. Every twin answers the same questions and sees every version. Twins compare the versions directly.
• Parallel — between-subject design, or A/B test. You test two to four versions, and each twin sees only one. No twin can compare them, so you learn how each version lands on its own.

When to use a single group

• You want a direct choice or a ranking — “Which do you prefer?” Show the versions together with the A/B Test or Rank Order question types.
• You’re comparing many versions, more than a parallel study can hold.
• There’s nothing to split — you’re exploring a topic, or testing one concept.

If the order in which twins see the versions could matter, randomise it.

When to use a parallel design

• Real people meet one version at a time. An ad in a feed or a pack on a shelf isn’t seen next to its alternatives. A parallel test mirrors that.
• Seeing one version would colour the next. Prices are the classic case: once a twin has seen the lower price, the higher one looks worse than it would on its own.
• The versions differ only subtly. Side by side, a small change — where the model is looking, one word in a headline — gets far more attention than it would in real life.
• You need a score for each version — appeal, purchase intent, understanding — rather than a winner.

Rule of thumb

“Which is better?” → single group. “How does each land on its own?” → parallel. When in doubt, a single group with randomised order is the safer default.

Setting up a parallel study

1. When you build a study from scratch, choose Parallel.
2. On the Study tab, add a condition for each version, up to four. Give each one a clear name, add its image, and optionally describe the scenario.
3. Write your questions once, below the conditions — every condition gets the same questions. If a question needs to differ for one version, open its Per-condition settings.

For the research background, see Research Design Basics.