Rank the list.
Build the set.

01 / Before fieldwork

Design

Paste the list and set the shape — items per screen, tasks and how often each item appears. Every item is seen equally often, and the interview stays under fifteen screens. Already have a design file? Drop it as-is, and the AI builds the design from it.

You get the design file to field, a report for the deck, and a sample size with the rule that binds it.

02 / After fieldwork

Analysis

Drop the fielded results — one row per respondent works too; it is converted for you. Hierarchical Bayes fits a score for every respondent, on CPU or GPU, and every run reports its checks:

• Chains agree (R̂), so you know the run settled
• In-sample hit for best and worst picks, against chance
• Most vs least tallies beside the model, as a model-free check
• 95% intervals on every share

Treat the order as the result; the shares are directional.

03 / Build the portfolio

TURF

The top five items often please the same people. Pick a portfolio size and TURF suggests the set that reaches the most of them. Pin or exclude items, see how each earns its place, how deeply people are covered, cut by segment, and compare against a benchmark — on the same study, without re-running the model.