What Is a Conjoint Question Type?
A conjoint question type is a survey question format that presents respondents with two or more hypothetical product profiles, each described by a set of attributes at varying levels, and asks them to choose which they prefer, rank them, or rate each one. The most common implementation is choice-based conjoint (CBC), where respondents see a series of choice tasks, each displaying 2-4 product concepts composed of different attribute-level combinations. By analyzing patterns across many choices made by many respondents, researchers calculate the relative importance of each attribute and the utility value of each level, revealing which product configurations drive preference and how respondents trade off features against each other. Conjoint questions are the survey-based engine behind pricing optimization, feature prioritization, and product line design.
Why Conjoint Questions Matter
Traditional survey questions ask about attributes in isolation, "How important is battery life?", which produces inflated importance scores because respondents don't have to make trade-offs. Conjoint forces trade-offs by embedding attributes within complete product profiles. A respondent can't say everything is important; they have to choose between a product with long battery life and high price versus one with shorter battery life and lower price. This mirrors real purchase decisions and produces preference data that's far more predictive of actual market behavior than direct importance ratings.
How Conjoint Questions Work
Anatomy of a Conjoint Task
Each choice task presents a set of product profiles displayed side by side. Each profile is defined by the same attributes (e.g., brand, price, battery life, screen size) but at different levels (e.g., Brand A vs. Brand B, $499 vs. $699, 8 hours vs. 12 hours). Respondents select the profile they'd most likely choose.
A typical CBC task looks like:
| Option A | Option B | Option C | |
|---|---|---|---|
| Brand | Apple | Samsung | |
| Price | $999 | $799 | $699 |
| Battery | 12 hours | 10 hours | 14 hours |
| Storage | 128 GB | 256 GB | 128 GB |
The respondent picks one. Many designs also include a "none of these" option to allow respondents to reject all profiles, which is important for realistic demand estimation.
Experimental Design
The attribute-level combinations shown in each task aren't random, they follow a structured experimental design that ensures all attributes and levels appear with balanced frequency and in uncorrelated combinations. This design enables the statistical model to estimate individual attribute effects cleanly.
Full-profile designs vary all attributes simultaneously across tasks. This is the standard for CBC and produces the richest trade-off data.
Partial-profile designs show only a subset of attributes per task, reducing cognitive load when the total attribute count is high (7+). Each task varies 3-4 attributes while holding others constant.
D-optimal designs algorithmically generate task sets that maximize the statistical precision of utility estimates for a given number of tasks. Most conjoint software uses D-optimal or near-optimal design algorithms.
Number of Tasks
Each respondent typically completes 8-15 choice tasks. Fewer tasks produce less precise individual-level utility estimates. More tasks increase fatigue, and response quality degrades after about 15-20 tasks for most respondents.
The optimal task count depends on the number of attributes, levels, and the estimation method. Hierarchical Bayes estimation, the standard for CBC, produces stable individual-level estimates with 12-15 tasks for designs with 4-6 attributes.
Attribute and Level Design
This is where most conjoint studies succeed or fail, and it happens before a single respondent sees the survey.
Attributes should be decision-relevant, independently meaningful, and actionable. "Build quality" is hard to operationalize in a conjoint task because respondents interpret it differently. "Warranty length" (1 year, 2 years, 3 years) is concrete and comparable.
Levels within each attribute must span a realistic range. If price levels don't include the range respondents would actually consider, the utility estimates for price won't reflect real market sensitivity. If all levels are too similar, the task doesn't create meaningful trade-offs.
Number of attributes: 4-7 is the practical range for full-profile CBC. Below 4, the choice tasks are too simple to generate interesting trade-offs. Above 7, cognitive overload sets in and respondents start using simplifying heuristics (like always choosing the cheapest option) rather than evaluating complete profiles.
Number of levels per attribute: 3-5 levels per attribute is standard. Unequal numbers of levels across attributes can bias importance estimates, attributes with more levels tend to appear more important than they actually are (the "number of levels effect"). Aim for balanced level counts when possible.
Respondent Experience
Conjoint tasks are more demanding than typical survey questions. Best practices for the respondent experience:
Introduce the task clearly. Before the first choice task, explain what respondents will see and what they should do. A brief instruction screen with an example task reduces confusion and improves data quality.
Keep the visual layout clean. Each profile should be easy to scan. Use consistent formatting, clear attribute labels, and enough white space that profiles don't blur together. On mobile, consider stacking profiles vertically rather than side by side.
Mix in engagement checks. A fixed-choice task where one profile clearly dominates (e.g., same features at half the price) serves as an attention check. Respondents who don't select the dominant option may not be engaging with the task.
When to Use Conjoint Questions
- New product design to determine which feature combinations maximize preference and willingness to pay
- Pricing strategy to estimate price sensitivity and optimal price points in the context of other product attributes
- Competitive positioning to understand how your product's attributes trade off against competitors' in the minds of buyers
- Product line optimization to identify which configurations to offer and which to retire
- Go/no-go decisions where you need demand simulation data to forecast market share for potential product configurations
Common Mistakes to Avoid
- Including too many attributes (8+) which overloads respondents cognitively and produces unreliable utility estimates, if you have more than 7 attributes, use partial-profile designs or split them across separate exercises
- Choosing unrealistic attribute levels that respondents would never encounter in market, utility estimates based on fantasy levels don't predict real purchase behavior
- Skipping the pilot test: conjoint tasks are complex enough that even small design issues (confusing labels, awkward layouts, unclear instructions) significantly degrade data quality; always pilot with 10-20 respondents
How Quali-Fi Supports Conjoint Questions
Quali-Fi's Research and Intelligence plans include choice-based conjoint with D-optimal experimental design, Hierarchical Bayes estimation, and built-in market simulation tools. The platform handles attribute and level configuration, generates optimized task sets, and produces individual-level utility scores, so you don't need to outsource to specialized conjoint consultancies.
Frequently Asked Questions
How many respondents do I need for a conjoint study?
For Hierarchical Bayes estimation (the standard), 200-300 respondents per segment is typical. Smaller samples (100-150) can work for aggregate-level analysis, but individual-level utilities become unstable. If you're comparing conjoint results across 3-4 segments, budget for 200+ per segment.
Can conjoint questions work on mobile?
Yes, but the layout needs adaptation. Two or three product profiles displayed side by side may not fit on a phone screen. Use responsive designs that stack profiles vertically or allow swiping between options. Limit to 2-3 profiles per task on mobile to maintain readability.
What's the difference between conjoint and MaxDiff?
Conjoint measures trade-offs across multiple attributes of a complete product profile. MaxDiff ranks items on a single dimension (importance, preference, appeal). Use conjoint when you need to understand how attributes interact and trade off. Use MaxDiff when you need to prioritize a list of features, messages, or concepts.
Related Topics
- Conjoint Analysis
- Conjoint Analysis Design
- MaxDiff Question Type
- Van Westendorp Question Format
- Questionnaire Design
Ready to run conjoint analysis without the complexity? Start a free trial of Quali-Fi Research and use built-in conjoint with automated experimental design and Hierarchical Bayes estimation.