Ask customers whether they want more storage, better support and a lower price, and agreement is easy. The interesting decision begins when they cannot have all three. Conjoint analysis creates those trade-offs systematically, helping researchers estimate how different elements of an offer contribute to preference.
It is a family of methods in which people evaluate product profiles made up of attributes and levels. For a software plan, attributes might include storage, support and price; the levels are the specific options within each attribute. In choice-based conjoint, participants repeatedly choose between profiles. Other forms use ratings or rankings, distinctions described in Sawtooth Software’s comparison of conjoint methods.
Build choices that could plausibly exist
The design of the study determines what the model can learn. Attributes need to cover the decision adequately without making each task overwhelming, and their levels need to be understandable and credible. A beautifully balanced statistical design is of little help if respondents cannot imagine using any of the offers.
Consider an illustrative delivery service study varying delivery window, collection option and price. If the lowest price always accompanies the slowest delivery, the analysis may struggle to separate their contributions. A considered experimental design varies the combinations so that the relevant effects can be estimated, while excluding combinations that genuinely cannot exist. Specialist design and analysis software can help, but it does not choose sensible attributes on the researcher’s behalf.
Interviews and a pilot study are useful preparation. They can reveal a missing consideration, ambiguous wording or tasks that demand more attention than the audience can reasonably give.
Read the model as a model
Analysis estimates utilities, sometimes called part-worths, for the attribute levels. These express relative preference within the fitted model. Attribute importance also depends on the range of levels tested: a study comparing very different prices may make price appear more influential than a study comparing a narrow range.
Simulations can explore how choices might shift between proposed offers, including competitors where represented. Their outputs are conditional on the alternatives, sample and assumptions. A simulated preference share is not automatically a forecast of market share; awareness, distribution, availability and the difficulty of switching may all affect real purchases.
For willingness-to-pay work, take particular care when translating utilities into money. The answer depends on the price model and competitive setting. Conjoint is valuable because it makes trade-offs explicit, but it still asks people to respond to a constructed decision. Subsequent purchasing evidence remains a useful test of what the model suggests.