How to Run a Conjoint Experiment
A complete high-level workflow for designing, fielding, analyzing, and documenting a conjoint experiment.
Quick Answer
To run a conjoint experiment, define the choice task, write attributes and levels, add restrictions only for impossible profiles, pilot the survey, field it by URL, and export profile-level data for AMCE or ATE analysis.
Design the task first
Start with the real choice respondents should make. Then define the profile unit, attributes, levels, choice prompt, task count, and whether respondents can choose neither profile.
Pilot before launch
A short pilot checks timing, comprehension, mobile layout, randomization, URL parameters, and export shape before the full sample budget is spent.
Analyze and document
After fielding, use long-format exports, respondent-level clustering, codebooks, methods text, and appendix materials to support publication or review.
FAQ
How many tasks should respondents complete?
There is no universal number. Use enough tasks for precision while keeping respondent burden reasonable, then pilot completion time and fatigue.
Should I use impossible-profile restrictions?
Use restrictions only for substantively impossible combinations, document them, and interpret estimates conditional on the allowed profile space.