Vignette Experiment Example

A vignette experiment shows each respondent a short profile — a candidate, policy, neighborhood, job, or message — then measures a reaction. In Domandata this is a conjoint question with one alternative, Display as Vignette, and optional random assignment around it. You do not need custom JavaScript to draw attributes or keep the shown profile in the export.

Suppose you want to know how people judge a local housing policy. Each respondent sees one paragraph that randomly combines cost, tenure, and who benefits. They then rate support on a Likert item. The analysis file needs the rating plus the three attribute levels that appeared in that paragraph.

• Attributes: monthly cost, tenure type, beneficiary group.
• Levels: at least two per attribute, with prohibited combinations if a pairing would never occur.
• Display: one alternative, vignette paragraph, selection turned off so the page is read-then-rate rather than choose-A-or-B.
• Outcome: a Likert or grid item on the next page.

1. Step 1: Add a Conjoint question. In Block Builder, choose Conjoint and keep a single alternative column. That column is the vignette, not a choice set.

   Step 1: Choose Conjoint question type

2. Step 2: Switch Display as to Vignette. Add attributes and levels, then under Display as choose Vignette. Turn off selecting an alternative so respondents read the profile instead of clicking it. Preview a few draws to check that the paragraph reads as a natural description.

   Step 2: Set vignette display

3. Step 3: Measure the outcome and export the profile. Place outcome questions after the vignette. Open Export after a pilot: the assignment snapshot still records which levels were shown, even though there is no selected-alternative click.

   Step 3: Preview the vignette draw

Use a multi-alternative table when the research question is a choice between profiles, not a reaction to one description. The same conjoint question type covers both; only alternative count and Display as change. See Conjoint Analysis Survey Design for tasks, prohibited combinations, and balancing across respondents.