Conjoint Rank upgraded to Hierarchical Bayes scoring model (HB-MNL) and HMC scenario simulator
This is the first in a series of improvements to OpinionX’s Conjoint Rank question type.
What Changed?
On December 9th 2025, we replaced OpinionX’s old linear-regression scoring algorithm with a Hierarchical Bayes Multinomial Logit (HB-MNL) model. Additionally, the scenario simulator chart now uses a Hamiltonian Monte Carlo (HMC) sampling system. These are best-practice, industry-standard approaches for choice modelling, so this update will give you more accurate, stable results from your same OpinionX survey data.
Why Does This Matter?
HB-MNL borrows strength across respondents, so individual preference scores are estimated more robustly, even if you collect significantly fewer votes than on the old LR scoring model.
HMC produces faster, more reliable sampling in the simulator, yielding better convergence and more accurate market-share forecasts.
Attribute importance is now reported as percentages that sum to 100%, making relative trade-offs clearer and attribute (category) importance a lot easier to interpret.
The Details:
Advanced Algorithm → Under the hood, HB-MNL applies a population-level prior and Bayesian estimation for each respondent. In effect, each person’s part-worth utilities converge more quickly and accurately.
Attribute Importance Chart → The importance of each category previously showed the part-worth utility range of each category’s highest and lowest-scored levels. The new HB-MNL model shows the proportion of total preference that participants allocate to this attribute (summing to 100% across all attributes). This percentage‐based chart is the same output style used in HB-based conjoint analyses, giving a clearer picture of how much each category drives choices along with a more industry-standard chart format.
Richer Participant Scores → Because of the HB approach, individual preference scores are much more stable and precise, even with fewer votes per respondent. In practical terms, you’ll find that each person’s utility profile feels more consistent and less noisy than before, reflecting real preferences more faithfully.
Improved Scenario Simulator → Our scenario simulator now runs its MCMC sampling via Hamiltonian Monte Carlo. HMC is a smarter sampler that explores the utility space more efficiently than traditional sampling. The result is faster convergence and more accurate market-share forecasts for any hypothetical profiles you create.
Automatic Roll‑Out → All existing conjoint surveys (free and paid) have already been re-scored with this new model, so nothing is required on your part. Your old results and charts were already updated to the improved HB analysis and recalculated automatically.
What you need to do
Nothing! This update has already happened and your old Conjoint Rank surveys have already been migrated to the new model. Your survey setup and participant voting experiences remain exactly the same as before. What you may notice is higher-fidelity results and more intuitive charts.