Conjoint Analysis

When a company builds a new product, one question comes before all the others: what do customers actually want? The answer sounds simple — just ask them. But asked directly, customers want everything, and they want it cheap.

Rate each feature on its own and the survey comes back the same way every time: better quality, five stars; lower price, five stars; faster delivery, five stars; more features, five stars. Everything is important. That result is useless to a team working with a fixed budget, because nobody can maximize every feature at once.

Why Rating Surveys Fail

Imagine launching a new pair of wireless headphones. Should you prioritize sound quality, battery life, a lower price, premium design, or free shipping? Put that question to customers directly — "which feature matters most?" — and the honest answer is "all of them."

That tells a product team nothing. In real life, customers compromise constantly: they accept a higher price for better sound, or trade battery life to save money. A rating survey never sees those compromises, because it lets people say yes to everything. Conjoint analysis recreates the decision itself instead of interrogating features one at a time.

What Conjoint Analysis Actually Does

Conjoint analysis is a market-research technique that estimates how much value customers place on each product feature — not by asking, but by watching them choose. Respondents don't rate attributes; they pick between complete product concepts, each a different bundle of features.

Instead of "which feature is important?", the respondent answers a far more realistic question: which one would you buy? Every choice leaks information about what they value. After hundreds of responses, the model works backward to estimate how much each attribute contributed to the decisions.

How It Works, Step by Step

1. Define the attributes and levels

Start with the characteristics customers weigh when buying — price, brand, storage, delivery, warranty, battery — and give each attribute a few concrete levels.

2. Build realistic choice tasks

Respondents see complete product profiles built from those levels and pick the one they'd actually buy. Because they choose a whole product, the task mirrors real shopping rather than an abstract feature quiz.

3. Estimate part-worth utilities

From the pattern of choices, the model assigns every level a part-worth utility — think of it as preference points. Higher utility means a level pulls choice upward; negative utility pulls it down.

Attribute Importance

Once every level has a utility, the range within each attribute — its best level minus its worst — tells you how much that attribute swings the decision. Normalize those ranges and you get attribute importance.

This is where conjoint earns its keep. The feature that topped a rating survey often isn't the real driver of choice — importance is measured by how far a feature actually moves decisions, not by how highly people say they like it.

Willingness to Pay

When price is one of the attributes, utilities can be translated into money. The price levels tell you how many utility points a dollar is worth; divide any feature's utility by that rate and you get its dollar value — its willingness to pay.

Simulating Products Before They Exist

The most powerful use of conjoint is simulation. Once utilities are estimated, you can assemble any hypothetical product — even ones you haven't built — and predict the share of customers who would choose it against competitors.

• What happens to share if we raise the price by $50?
• Which feature can we drop to cut cost with the least damage?
• Which configuration maximizes market share?
• How does a competitor's new model pull demand away from ours?
• Which version should we launch first?

Each of these becomes a scenario you can compare with numbers instead of instinct — before spending on production, inventory, or marketing.

Where Conjoint Is Used

Common Mistakes

A well-run conjoint study is powerful; a careless one produces confident-looking nonsense. Most failures trace back to a survey that asks too much of respondents.

• Testing too many attributes, so respondents tire and choices turn random
• Including feature combinations that could never exist in the real market
• Using attributes customers genuinely don't care about
• Making the survey too long — fatigue quietly corrupts the data
• Reading part-worths as absolute scores instead of relative preferences

Conjoint in the SKARI Survey Lab

SKARI's Survey Lab runs the full conjoint workflow — from experimental design to business interpretation — without code or separate statistical software. It supports every major variant, so you can match the method to the study.

• Choice-Based Conjoint (CBC), Adaptive (ACA), Adaptive CBC (ACBC), and rating-based conjoint
• Automatic, balanced experimental design for the choice tasks
• Part-worth utility estimation with bootstrap confidence intervals
• Attribute importance overall and re-estimated by segment (a per-segment heatmap that shows who is price-driven vs brand-driven)
• Willingness-to-pay conversion when a price attribute is present
• A market-share simulator for what-if products and competitive scenarios
• An auto-generated executive summary that reads the results in plain language

Frequently Asked Questions

Is conjoint better than a rating survey?

For pricing and product design, clearly yes. Rating surveys make every feature look important because respondents judge them in isolation. Conjoint forces trade-offs, so the priorities that matter actually separate.

How many attributes should I include?

Most studies use four to six attributes, each with two to five levels. Beyond that, respondent fatigue sets in and data quality falls.

Can conjoint really estimate willingness to pay?

Yes — as long as price is one of the attributes. The price levels calibrate utility into dollars, so any feature's utility converts to a monetary value.

How many respondents do I need?

Many choice-based studies use 200–500 respondents, though the exact number depends on how complex the experimental design is.

Key Takeaways

Conjoint analysis answers the question every product team is really asking: what do customers value enough to choose? Rather than asking people what they like, it observes what they pick when realistic trade-offs are on the table.