Conjoint
Quali-Fi’s Conjoint question is choice-based conjoint (CBC). Each task shows a set of profiles built from the attributes you define. The respondent picks one profile — or rejects all of them if you include a none option.
There is one conjoint question type. It is not ACA, ACBC, or MaxDiff. For most/least item scoring, use MaxDiff. For price-only sensitivity, use Van Westendorp or Gabor Granger.
Note
Conjoint appears under Matrix & Structured Evaluation in Add Question. If it is missing from your account, contact your Quali-Fi account team — the type is package-gated.
Display modes
In Properties → Conjoint Settings, Display mode is one of:
| Mode | What respondents see |
|---|---|
| Classic CBC cards (default) | Side-by-side cards labeled A, B, C (or your custom labels). Each card lists every attribute as a labeled row. |
| E-commerce shopping shelf | Product-style cards with image, title, price, badge, rating, and a full-width select button. |
Display mode changes presentation only. The stored answer is still “which alternative (or none) was chosen on each task,” plus the attribute levels shown on those cards.
Existing surveys without a display-mode setting render as classic cards.
How to create
Open your survey in the Survey Builder.
Click Add Question, open Matrix & Structured Evaluation, and select Conjoint (“Evaluate multi-attribute trade-offs”).
Enter the question prompt — the comparison instruction above the tasks.
In Properties, configure Conjoint Settings. New questions start with:
- 8 tasks per respondent
- 3 alternatives per task
- Randomize tasks on
- Show progress on
- Include none option off
- Two starter attributes: Brand (Brand A / Brand B) and Price ($9.99 / $14.99)
Under Attributes, rename the starters, add levels, and add more attributes. Each attribute needs at least two levels.
If you want a shopping-shelf layout, set Display mode to E-commerce shopping shelf. Quali-Fi adds a Product Image attribute when none exists, and it will assign Title / Price card roles to attributes whose labels look like Brand or Price.
Preview a full run. The canvas preview is labeled Choice-based conjoint preview or E-commerce conjoint preview. Then complete the live Preview flow to confirm task navigation.
Warning
The respondent runtime needs at least two attributes, and each usable attribute needs at least two levels. Anything thinner shows: “Please add at least 2 attributes with 2 levels each to use Conjoint.” Attributes with fewer than two levels are dropped at answer time.
Conjoint Settings
These controls sit in Properties under Conjoint Settings.
| Setting | Default | What it does |
|---|---|---|
| Display mode | Classic CBC cards | Classic cards or e-commerce shopping shelf |
| Tasks per respondent | 8 | How many choice tasks each person completes. Minimum 1. Clearing the field and leaving it restores 8. |
| Alternatives per task | 3 | How many profiles appear in each task. Minimum 2. Clearing the field and leaving it restores 3. |
| Randomize tasks | On | Builds a deterministic set of profiles from the question. The same settings produce the same tasks for every respondent. This is not a unique design per person, and it is not a formal experimental-design engine. |
| Show progress | On | Shows {{completed}} of {{total}} complete (for example, “3 of 8 complete”) |
| Include none option | Off | Adds a reject-all control under the cards |
| None option label | None | Shown only when none is on. In e-commerce mode, turning none on while the label is still None changes it to Exit store without making a purchase. |
There is no upper bound in the editor for task count or alternatives. Keep both small enough that people can finish. Survey Health treats conjoint as a high-burden question (about 8–18 seconds per task) when quality-control estimates are on your plan.
The footer under Attributes reports how many attributes (and prohibitions, if any) are configured. Properties does not include a prohibition editor. See Forbidden combinations.
Alternative labels
Classic cards use letter labels by default: A, B, C. The question can also use Option 1 / Option 2 or custom strings. Those label modes are not in Properties. If your package includes Survey Script Mode, you can set them there (conjoint label_mode and conjoint alternative_labels). Otherwise the live question stays on letters.
Attributes and levels
Each attribute is a factor in the profiles (Brand, Price, Warranty). Each level is a value of that factor.
- Add attribute creates
Attribute Nwith Level 1 and Level 2. - Add level appends another level on that attribute.
- You can rename the attribute and every level.
- Remove on a level is disabled when only two levels remain.
- You can remove an entire attribute.
Keep labels short. Respondents scan several cards at once.
Card roles
Every attribute has a Card role. Roles mainly change the e-commerce card. Classic cards still list every attribute as a row; a Rating attribute shows stars instead of the level text.
| Role | What it maps to | Exclusive? |
|---|---|---|
| Plain (default) | Detail row (Warranty: 2 years) | No |
| Image | Product photo | Yes — one attribute |
| Title | Product name under the image | Yes |
| Price | Prominent price text | Yes |
| Badge | Promo overlay on the image | Yes |
| Rating | Stars under the title | Yes |
Assigning an exclusive role that is already in use shows: “Image role is already assigned to another attribute.” Unknown roles render as Plain.
Rating attributes
When the role is Rating:
- Max stars is 3–10 (default 5).
- Each level has a star value you set in the builder.
- Respondents see the stars. They do not pick a rating; the rating is part of the profile.
Product images (e-commerce)
In e-commerce mode, Product images appears above the attribute list. Upload one image per level on the image-role attribute.
- Prefer Upload image / Browse. That stores a durable file id. Quali-Fi serves it through a public file URL (S3 signed URLs are refreshed so stored survey links do not expire).
- You can paste an Image URL. Do not paste a temporary S3 signed URL — those expire. The builder warns when a level looks like an expiring Amazon link.
- Clear image removes the file from that level.
- If no image attribute exists, Add image attribute creates Product Image with Product A / Product B.
Image URLs live on the question settings. They are not written into the answer payload. Cards resolve photos from the configured level_id at render time.
When none is still off in e-commerce mode, the builder suggests turning it on so people can leave the “store” without buying.
E-commerce-only copy
These fields appear only when Display mode is E-commerce shopping shelf:
| Setting | Default | What it does |
|---|---|---|
| Select button label | Add to Cart & Checkout | Full-width CTA on each product card |
| Task prompt | (empty) | Optional shopping-trip line above each task |
| Hide alternative labels | On | Hides A/B/C on the product cards |
What respondents see
Respondents work one task at a time.
- Numbered task buttons appear when there is more than one task. A check mark marks a completed task. They can jump back.
- Choosing a card (or none) saves that task and moves to the next task.
- Show progress prints how many tasks are done.
- A required conjoint question is not complete until every task has a card or a none selection.
Classic cards: letter (or custom) label, then one row per attribute. Selected card uses a violet highlight.
E-commerce cards: image (or a “No image” placeholder), optional badge, title, stars, price, plain-attribute details, then the select button. Selected card uses an emerald highlight. The none control, when enabled, sits centered under the shelf.
If you change attributes, task count, or alternatives after someone has already stored answers, those stored task profiles stay with the response. New respondents get the current generated set.
Validation
Frontend and backend both enforce a complete conjoint payload.
Required questions
- Every configured task needs a selection.
Nonecounts as complete when Include none option is on.- Incomplete required tasks show “This question is required.”
Submit checks (respondents do not see these codes; they matter when an answer is edited or imported)
- Fewer tasks than configured on a required question → rejected
- Fewer than two alternatives on a task → rejected
- Duplicate alternative ids, unknown attribute/level ids, or the same attribute twice on one card → rejected
- None selected when none is off → rejected
- None plus a chosen card on the same task → rejected
- Chosen alternative that is not in that task → rejected
Optional conjoint questions can be left unfinished.
Survey Health counts conjoint as a high-effort task for length estimates. It does not score individual conjoint answers.
Rhoxi survey simulations do not invent free-text conjoint choices. They build a complete task set from your settings and pick a profile on every required task, varying selections across simulated respondents so the question can submit.
Forbidden combinations
The runtime can refuse a profile that pairs two specific levels (if Brand = X, then Price cannot be Y). Generation tries up to 20 draws per card, then falls back to the first level of each attribute.
Properties shows the prohibition count. It does not let you add or edit them.
If your package includes Survey Script Mode, you can author a line such as:
conjoint prohibition: Brand=Brand A -> Price=$9.99That means “do not show Brand A and $9.99 on the same card.” Without Script Mode, leave prohibitions empty.
Script Mode extras
Survey Script Mode is a separate package feature. When it is on, the script can also set display mode, task counts, none, e-commerce copy, attribute roles, image metadata, rating values, label mode, and prohibitions. Switching back to Visual mode shows those settings in Properties where the panel supports them.
Script Mode is the only builder UI for prohibitions and alternative-label mode.
Data, analysis, and exports
Analytics, the Conjoint Workspace, scenario simulation, Raw Data columns, and CSV/SPSS layouts are covered in Conjoint Analysis and Data Layouts.
Short version: there is no single summary column. Each task expands into choice, choice position, none flag, and the level shown on every alternative × attribute.
Best for
- Packaging and feature trade-offs
- Pricing work where brand, features, and price have to compete on the same card
- Studies that need utilities and market-share scenarios after fielding
Use MaxDiff when you only need most/least item priority. Use Swipe Cards when people should react to one stimulus at a time, not compare full profiles.
Tips
- Two to four alternatives per task is the usable range. More cards than that get skimmed.
- Eight tasks is a starting point, not a target. Add tasks only if the design needs them.
- In e-commerce mode, upload images through Browse. Pasted signed URLs will die in field.
- Preview the whole sequence. Task buttons, none, and the e-commerce CTA are easy to get wrong on the first pass.
- Do not treat the generated cards as a published experimental design. If you need a specific orthogonal plan, build the attributes carefully and review the preview tasks — Quali-Fi is not running a separate design-of-experiments engine.