Conjoint Analysis

Paul Green, a professor at Wharton in the US is the father of conjoint analysis. In 1975, in the Journal of Consumer Research, he wrote an article with V Srinivasan entitled "Conjoint Analysis in Consumer Research: Issues and Outlook". It explained how market research can be used to determine how people value different components of an offer. It was a scientific approach to something that had previously been determined by psychology.

When people choose a product or service their decision is usually a trade-off of different factors. They balance the quality of the products against its price and the different benefits it will provide. A change to one of these attributes will change the likelihood of purchase. Conjoint analysis uses a mathematical program to find which attributes make the best product, what price should be charged, and what share of the market will be achieved. Conjoint analysis is a tool that predicts what people like and dislike about products and what will prompt them in their purchase choice. It is a "trade-off" tool that determines the most and least important factors in this decision-making process.

Conjoint analysis requires specialist software and in this, Sawtooth has a strong position. Conjoint analysis is best carried out by someone with a good statistical background and preferably somebody who has been trained in the use of the specialist software.

​

In the example below, a car is shown together with various features (attributes) and a price. The potential customer is asked to review the offer and say how likely they would be to buy the car on a scale from 1 to 7 (a 1 to 10 scale could just as easily be used) They would then see as many as 20 to 30 variations of this offer at different prices and asked to say their likelihood of purchasing it. When 200 or more people have been asked these questions, the software program carries out the trade-off analysis which determines the optimum price and mix of attributes. It also can show the value attached to each of the attributes and the share that the product could achieve if these were offered at a specific price.

The great attraction of conjoint analysis is the ability to arrive at a scientific assessment of what people value. However, there are limitations to conjoint which need to be borne in mind:

Sample size – In order to obtain reliable results it is usually necessary to interview at least 100 target respondents and preferably 200. In some business to business markets it isn’t possible to achieve these sample sizes.

Number of attributes – The design of the conjoint concept is critical. Conjoint analysis becomes unreliable if there are a large number of different attributes, including price. Similarly, the number of variables for each attribute needs to be limited to between three and five levels, otherwise the potential combinations of concepts becomes too large to manage.

Respondent fatigue – Respondents who take part in a conjoint survey are asked to look at a number of different screens, each with different offers (with different prices). It is easy for them to become “punch drunk” with so many offers to view. Unless the attributes and the variables are quite distinctive, the offers can blur and seem very similar. When this happens, respondents get confused and tired and do not give their full consideration to the choices, choosing any at random to complete the tedious interview.

Some things to think about:

• Conjoint is appropriate if you are trying to work out the importance of the different mix of attributes in a product. However, if all the attributes are fixed and locked in, and you just want to test prices, it might be better to use Van Westendorp or Gabor Granger - https://www.b2bframeworks.com/frameworks/new-product-pricing.

• When designing your conjoint survey, limit yourself to 5 or 6 different attributes plus price. If you have too many attributes, you need a large sample size and it can be confusing to respondents if they are shown too many variations of the offers.

• If you can only find a maximum of 100 respondents to take part in the survey, consider using a different tool such as SIMALTO - https://www.b2bframeworks.com/frameworks/simalto.