Best-worst scaling in transport research
Best-worst scaling produces relative importance rankings across a large attribute set with less respondent burden than a DCE. It does not produce WTP estimates.
What best-worst scaling is, when it is appropriate in transport research, and what it produces compared with a DCE.
Knowledge Base -> Experiment Design -> Transport
Ben White, 07.07.2026
When a DCE is not the right tool
Best-worst scaling asks respondents to identify the best and worst options from a presented set. Applied to transport attributes, it reveals which attributes are most and least important across a large candidate set - more attributes than a DCE can efficiently accommodate.
BWS does not produce WTP estimates or utility parameters. It produces relative importance scores - which attributes matter most but not how much they are worth in monetary terms.
Types of best-worst scaling in transport
BWS is appropriate for attribute screening - identifying which of a large candidate set to include in a DCE. A BWS study with a much large number of attributes can be conducted with the same respondent burden as a much smaller DCE
In some transport policy contexts, relative importance rankings are the required output. Passenger satisfaction research and service quality prioritisation exercises often require importance rankings rather than monetary values.
TLDR Quick links
Setting up best-worst scaling in SurveyEngine
Step 1: Create one Attribute with each level as the 'atomic' items to be evaluated. Then create 3 alternatives (at least), these are what best and worst are chosen from.
Step 2: Generate a balanced incomplete block design using the standard design generator in SurveyEngine. Each respondent sees a subset of items. The design ensures each item appears the same number of times.
Step 3: Calculate best-worst scores: best choices minus worst choices for each item, standardised by appearances.
Step 4: Use BWS results to select the highest-scoring attributes for the subsequent DCE.
Step 5: Do not confuse BWS scores with WTP estimates. They are ordinal relative importance measures only.
Worked example - transport service quality ranking
A national rail operator needs to identify which of 20 service quality attributes to include in a SP study. A BWS study with 500 rail users ranks all 20. Punctuality, seat availability, and journey time reliability score highest; station retail and WiFi score lowest.
The top 6 attributes by BWS score are included in the subsequent DCE for WTP estimation.
References
Flynn, T.N. et al. (2007). Best-worst scaling: What it can do for health care research. Journal of Health Economics, 26(1), 171–189.
Ready to run a best-worst study? Log in to SurveyEngine and follow the best-worst tutorial.
Or Contact us at support@surveyengine.com — we're glad to help.