A MaxDiff study lives in the Trade-off Platform. Its header shows the study name and three steps — Design, Analysis and TURF — each unlocking when the one before it is done.
Design the survey — or upload yours
Items. List what the study ranks, one per row. Paste a whole list into any row and it splits into items. Three is the minimum, 30 the most you can generate a design for (60 if you upload one), and 8–20 is the sweet spot.
Already have a design file? Drop it as-is, and the AI builds the design from it.
Generate a design. Set Options (items per screen, 3–6), Tasks (1–15) and Exposure (how often each item is seen, 2–5); a line under them says what that produces, such as "Each item is seen 2.4× across 6 tasks of 4." Press Generate design.
When the design is ready you get:
- Design file ready — screens × items each, the respondents to field, and checks such as "Every item seen equally often" and "Every item pair compared".
- Sample size — how many respondents to field, and what changing it costs. Edit the number and the design adjusts; every framework is shown, and the one that binds is marked.
- Design preview — what respondents see, screen by screen.
- Downloads: the design file to field (CSV) and the design report (PDF).
Analyse the results
Drop your fielded results. One row per respondent × task × shown item, or one row per respondent — the second is converted for you. Upload it as an SPSS (.sav) file.
Columns. Respondent, Task, Item, Best and Worst are auto-matched from the file; check them before running. You can add segment columns too.
Estimation. Hierarchical Bayes fits one set of scores per respondent. Choose a preset — Fast draft, Standard or Thorough — or set Warm-up draws and Kept draws, choose to Run on CPU or GPU, and press Run the analysis. It keeps running if you close the tab.
The model. When it settles you see What matters most — each item's share of total importance, with 95% intervals — Most vs least important, the model-free tallies beside the ranking, and Study & model fit, including whether the chains agree (R̂) and the in-sample hit rate against chance. Treat the order as the result; the shares are directional.
Build the TURF portfolio
TURF opens on the scores you just fitted — no re-estimating.
- Portfolio size — pick how many items you can build, and press Suggest best set.
- Measured as — what counts as reached (an item in their top 1, top 3 or top 5) and which panel: everyone, or a segment.
- Items — tick items in or out, pin an item into every suggestion, or exclude one.
- Benchmark — freeze a portfolio, then read every change against it.
The results show the reach — how many respondents have at least one chosen item in their top N — How each item earns its place, the incremental ladder, and Depth of cover, how many of each respondent's favourites the set includes. Save the portfolio, export it as CSV, or download the deck.