Conjoint Analysis

Conjoint analysis is a survey method that works out how much each feature of a product is worth by forcing people to choose between whole bundles. It measures trade-offs, not opinions. You learn what buyers would give up, and what they would pay.

Nobody can reliably answer “how important is price to you?” So a conjoint survey shows a respondent two or three complete packages and makes them pick one — over and over, with the features shuffled each time.

From those forced choices you can work out the weight every attribute carried. The maths recovers what the respondent could not articulate, and often would not admit.

Key takeaways

• Conjoint analysis quantifies how much each product attribute is worth by forcing choices between complete bundles.
• Concept testing asks whether the idea lands; conjoint analysis asks what each part of it is worth.
• Typical outputs are part-worth utilities, attribute importance scores and a willingness-to-pay estimate.
• Regulators accept formal preference studies as evidence, so the method reaches well beyond marketing teams.

How it works

You define the attributes and the levels each can take, generate a set of bundles, then show respondents a handful at a time and record which one they pick. The pattern of those choices does the rest.

Run enough choice sets and the output reads like a price tag on every level: what six extra hours of battery is worth in dollars, and which group of buyers cares.

The analysis is not a spreadsheet job. A trained data analyst fits a choice model to the responses, and the survey design matters just as much as the model behind it.

Descriptive analytics tells you what customers already did — conjoint tells you what they would do when something is taken away from them.

Pricing teams use the results directly. A willingness-to-pay curve is what makes outcome-based pricing arguable in a negotiation rather than merely aspirational.

Bundle design follows on. If warranty turns out to be cheap to give and heavily valued, attaching it lifts average order value (AOV) without touching the headline price.

Fieldwork and modelling are often split. Data science outsourcing covers the modelling half when a company runs one study a year and cannot justify a permanent team.

Regulators take preference measurement seriously. The U.S. Food and Drug Administration (FDA) runs a patient-preference programme on the basis that only patients who live with a condition understand their care choices.

Its device centre gives that preference information a formal place in benefit-risk decisions, which is a long way from asking shoppers to rank features on a five-point scale.

Public-sector research guidance points the same way. The UK’s Service Manual carries user research through every design phase, from discovery to live, which is where preference work belongs rather than bolted on at the end.

Design the survey before you worry about the model. Too many attributes and respondents start guessing, so most studies cap the list at six or seven and test the rest another time.

Watch what people actually trade away, not what they claim to value. A respondent who never once gives up warranty has told you more than any importance rating could.

Examples

Conjoint has been in commercial use since the 1970s, when marketing academics adapted it from mathematical psychology. It shows up wherever a buyer has to accept less of one thing to get more of another.

Telecoms operators use it to design tariffs, testing data allowance against contract length, handset subsidy and monthly price until one plan wins on its own merits rather than on habit.

Medical device makers run formal preference studies to show which risks patients will accept in exchange for which benefits, then submit the results as evidence of what those patients genuinely value.

Airlines trade seat pitch, baggage allowance and change fees against the headline fare, which is how a carrier works out that extra legroom is worth more to a business traveller than a free checked bag.

Grocery brands test pack size, ingredient claims and shelf price together — a claim that wins in isolation often loses once the price rises to pay for it.

Software firms use it on packaging tiers, checking whether a feature belongs in the entry plan or is worth enough to hold back for the tier above it.

Employers have borrowed it for benefits design — asking staff to choose between pay, leave and flexibility packages tends to produce a very different answer from an engagement survey.

Related terms

Conjoint analysis sits among the quantitative research and pricing terms. Two of the neighbours describe the analytics work around it, one covers the pricing model it supports, and two name the people who own the modelling.

• Data Analyst: the role that fits and interprets the choice model behind the results.
• Descriptive Analytics: the backward-looking summary of what customers have already done.
• Outcome-Based Pricing: the pricing model that charges for results rather than effort.
• Customer Satisfaction Index: the tracked score summarising how satisfied a customer base is.
• Data Science Lead: the senior owner of modelling standards across a research team.

FAQ

What does conjoint analysis actually measure?

It measures the relative value of product attributes by observing real choices rather than asking for ratings. The output is a set of utilities you can convert into attribute importance scores and willingness-to-pay estimates.

How is it different from concept testing?

Concept testing asks whether an idea lands with buyers at all. Conjoint analysis assumes the idea is viable and works out what each part of it is worth.

How many respondents do you need?

Three hundred is a common working floor for a study with several attributes, and you need considerably more if you want reliable results for separate buyer groups. Sample design matters more than raw size.

What is a part-worth utility?

It is the value a respondent places on one level of one attribute, such as a three-year warranty against a one-year warranty. Utilities are relative, so they only mean something inside the study.

Is conjoint analysis only for pricing?

No, it also guides feature selection, bundle design, product portfolio choices and formal regulatory submissions.

Shortlist analytics partners with choice-modelling experience through the Outsource Accelerator directory.