What is Conjoint Analysis?

Conjoint analysis is a family of survey methods in which respondents evaluate product profiles made up of several attributes at different levels (a plan with 5 seats, unlimited storage, email support, at $29; another with 10 seats, 100GB, chat support, at $49) and their choices are analyzed to estimate the value, or part-worth utility, each attribute level contributes. Because respondents trade features against each other and against price, the method reveals what they would sacrifice for what, which direct questions ("how important is storage?") never do: asked directly, everything is important; asked to choose, priorities appear. The technique originated in mathematical psychology and marketing research in the 1970s (Paul Green's work at Wharton is usually cited as the foundation), and it has become the standard tool for pricing research, product configuration, and packaging decisions wherever the stakes justify a proper study.

The Main Variants

Choice-based conjoint (CBC). Respondents pick one profile from a set of two to four, repeatedly across many sets; the dominant form, because choosing mimics buying and the analysis (typically hierarchical Bayes estimation) recovers individual-level utilities from the pattern of choices.

Adaptive conjoint. The questionnaire adapts to each respondent's earlier answers, focusing on the attributes they seem to care about; useful for large attribute sets, at some cost in comparability.

Rating- and ranking-based conjoint. Older forms where respondents rate or rank full profiles; simpler to analyze, less realistic than choice.

Menu-based conjoint. Respondents build their own bundle from priced options, mirroring configurators and add-on pricing.

What It Delivers

Attribute importance. Which features actually drive choice, in proportion, across the sample and by segment.

Part-worth utilities. The value of each level of each attribute, which lets the team compare configurations that were never shown directly.

Price sensitivity. The utility of price levels, and therefore how much of a feature's value survives a price increase, the rigorous cousin of willingness-to-pay questions.

Market simulation. The most used output: a simulator that predicts share of preference for any set of competing configurations, so the team can test "what if we drop the free tier and add chat support to the mid plan?" against modeled competitors before doing it.

Designing One Well

1. Choose attributes and levels from research, not from the roadmap.
Five to seven attributes with two to five levels each is typical; they must be the things buyers actually weigh, discovered through interviews and exploration, and described in language buyers use.

2. Keep profiles realistic.
Combinations that couldn't exist, or prices outside the plausible range, produce utilities for a fictional market. Prohibit impossible pairings in the design.

3. Sample the buyers, at size.
Conjoint needs a few hundred respondents for stable estimates, from the population that buys, screened for role and relevance; recruiting a decision-maker sample through a screened panel (the kind a Ballpark study draws on) is usually the practical route.

4. Mind respondent burden.
Twelve to fifteen choice tasks is a comfortable ceiling; beyond it, fatigue degrades the choices the whole method rests on.

5. Validate with holdouts and behavior.
Hold back some choice tasks to check the model predicts them, and compare simulator predictions to real purchase data where any exists.

The Limitations

Conjoint is still stated preference: choices in a survey are not purchases, and hypothetical bias persists, if less than in direct questions. It can only value attributes it included, at levels it showed, so a missing attribute is invisible. It assumes respondents evaluate profiles rationally and attribute by attribute, which brand, habit, and inertia violate in real markets. And it is expensive in design skill and sample; for a quick read on a feature's appeal, lighter methods (MaxDiff, Kano, a preference test) come first, and conjoint is reserved for the pricing and packaging decisions that justify it.

Where This Leaves You

Conjoint analysis infers what people value from what they choose, trading features and price against each other in realistic profiles, and delivers importances, utilities, price sensitivity, and a simulator for decisions not yet made. Build the attributes from research, keep profiles real, sample the buyers at size, respect fatigue, and validate. It is the closest a survey gets to watching a purchase, and it earns its cost exactly where the decision is priced in revenue.