advanced

Conjoint Analysis and Data Layouts

Read conjoint utilities, run scenario simulations, and export task-level Raw Data, CSV, and SPSS columns

12 min readUpdated September 17, 2026

Conjoint Analysis and Data Layouts

Once respondents finish a Conjoint question, Quali-Fi estimates level utilities, shows task-level preference, and can simulate share for product mixes you save. The same task payload drives Raw Data, CSV, and SPSS.

Note

Screenshots for the Conjoint Workspace are not yet available in the Support site.

Where the numbers live

On a survey that includes a conjoint question, Analytics has three related surfaces:

  1. Overview → Question analytics — a conjoint card on the main analytics page
  2. Conjoint Workspace — a dedicated page for charts, scenarios, and workspace notes
  3. Report — the same conjoint visual on a slide (respondents, task completion, utilities)

Open the workspace from the analytics overflow menu (Conjoint Workspace) when the survey has at least one conjoint question.

The workspace loads the first conjoint question in survey order. Overview and the report builder still show a card for each conjoint question. If you put more than one conjoint question in a survey, treat the workspace as the first question’s analysis unless you have a reason to keep only one.

Submitted responses feed analysis. The overview card follows Include test responses and the date filter on Analytics. The Conjoint Workspace does not expose those filters — it uses submitted completes and excludes test responses. The zero-response banner appears until those completes exist.

Overview card

The question-analytics conjoint card shows:

TileMeaning
RespondentsPeople with conjoint answers in the current analytics scope
Completed TasksTasks with a chosen card or none
Completion RateCompleted tasks ÷ all stored tasks
Utility EngineHB — hierarchical Bayes–style utilities

Under the tiles:

  • Task Preference Summary — each task’s alternatives with win share. The top alternative’s share is called out. The list shows up to four alternatives per task.
  • Utility Summary — per attribute, Importance (percent) and Range, then each level’s centered utility (green when positive, red when negative).

The same two blocks appear in the report builder as the dedicated conjoint visual. Report chart-color palettes do not recolor those bars.

Conjoint Workspace

The workspace title is the survey name. The subtitle is the question number (Q3, or your custom number). Save in the top bar stores the workspace name, description, and a snapshot of the current analysis.

If the survey has no conjoint question (or no usable analysis yet), you see No conjoint question found.

When there are zero completed responses, a banner says analysis and scenario simulation need completed conjoint tasks before utilities or market shares can be estimated.

Summary row

TileMeaning
RespondentsCompleted responses used for utilities
AttributesAttribute count, plus total levels
TasksChoice tasks in the analysis
Top AttributeHighest-importance attribute and its percent

Analysis tab

  • Task Preference Summary — horizontal bars of the top alternative share by task
  • Attribute Importance — horizontal bars of relative importance
  • Attribute & Level Details — table of every configured attribute and its levels (the design, not the utilities)

Importance is the attribute’s utility range as a share of all attribute ranges. The first configured level of each attribute is the baseline in the HB-style fit (raw utility 0 before centering). Centered utilities then sum to zero within the attribute.

The engine label in reports is hierarchical-bayes (version 1.0.0). That is Quali-Fi’s built-in HB-style estimator — not an export to an external conjoint package.

Scenarios tab

Build a competitive set, save it, then simulate how respondents would split among those products.

  1. Name the scenario (default Scenario 1). Description is optional.
  2. Each product starts as Product A, Product B, and so on. Rename it. Pick one level per attribute.
  3. Add Product appends another mix. You cannot delete the last remaining product.
  4. Save & simulate (or Save Scenario when there are no responses yet) stores the mix. If respondents already exist, Quali-Fi runs the simulation immediately.

Saved Scenarios lists each mix. From a row you can:

  • Load — put that mix back in the builder
  • Run simulation / Re-run — refresh shares from current utilities
  • Delete the scenario

Each saved row shows Share of preference and First choice bars, respondent count, and the last simulation time.

When two or more saved scenarios have results, Compare scenarios lists share of preference and first-choice share for every product across those scenarios.

Workspace tab

Workspace Settings stores:

  • Workspace Name (default Conjoint Analysis)
  • Question (read-only prompt)
  • Description

This is notes and a cached analysis snapshot. It does not change the questionnaire.

How simulation scores a mix

Run simulation scores the saved mix with the HB respondent-level utilities already computed for the question.

MethodRule
Share of preferenceSoftmax over each respondent’s product utilities, then average
First choiceEach respondent goes to the highest-utility product. Exact ties go to the earlier product in the saved set

Shares describe only the products in that scenario. They are not a market forecast. There is no awareness, distribution, pricing elasticity, or outside good in the model.

Warning

The none / opt-out alternative is not simulated. Utilities do not include a none parameter. When the question has Include none option, the results still show: “None/opt-out is not simulated because respondent utilities do not include a none parameter.”

Re-run simulations after more completes arrive. Cached shares do not update on their own.

Empty utilities return an error. The workspace explains that on the zero-response banner.

Raw Data and exports

In Analytics → Raw Data, CSV, and SPSS, a conjoint question expands. There is no parent summary column in those tables.

If a leftover parent cell is rendered somewhere, it may still read like task 1: A | task 2: none. That string is a display leftover. Use the expanded columns for analysis.

Column pattern

Headers use the question number (for example CONJ or Q5) plus a task suffix. Attribute keys are slugified from the attribute label (Meals available → meals-available). Alternative keys come from the card labels (A → a, Option 1 → option-1).

For each task tN:

ColumnWhat it holds
{Q}_tN_choiceChosen card label (A, B, …) or none
{Q}_tN_choice_position1-based position of the chosen card. Empty when none or no choice
{Q}_tN_none1 if none, otherwise 0 (this column is always exported)
{Q}_tN_{alt}_{attribute}Level shown on that alternative for that attribute — one column per alternative × attribute

Example for question CONJ, 3 alternatives (A/B/C), attributes Brand and Price, task 1:

CONJ_t1_choice
CONJ_t1_choice_position
CONJ_t1_none
CONJ_t1_a_brand
CONJ_t1_a_price
CONJ_t1_b_brand
CONJ_t1_b_price
CONJ_t1_c_brand
CONJ_t1_c_price

The shown-level columns record the design the person saw, not only the winning card. That is what you need to rebuild the task.

Values and answer format

Choice and none stay as labels / 0–1. Shown levels follow the export Answer format:

Answer formatShown-level cell
Label only (typical Raw Data default)Brand A
Code onlythe level id (brand_a)
Code and labelbrand_a: Brand A

CSV uses the same hyphenated headers as Raw Data.

SPSS uses the same columns and values. SPSS identifiers cannot contain hyphens, so attribute keys become underscores (CONJ_t1_a_meals_available).

Editing an answer

If you can edit a raw conjoint answer, the editor reuses the live task cards. Saving still has to pass the same completeness and id checks as submit. Do not change the stored alternatives unless you intend to rewrite the design that respondent saw.

What this is not

  • Not a Sawtooth / external HB export. You get Quali-Fi’s utilities and scenario shares, plus the task-level file.
  • Not a per-respondent randomized design file. Tasks are generated from the question settings; see Conjoint.
  • Not a none-inclusive market model. Opt-out is collected on the question and exported; it is not part of simulated share.
  • Not piping. Conjoint answers are not offered as recall tokens the way choice questions are.

Related

Was this article helpful?