Posted on 28 December 2016
Conjoint analysis is a very powerful tool for market research. Here are Conjoint.ly’s best practice tips for setting up your analysis to yield useful results.
- Whilst it’s tempting to put as many attributes and levels as possible, this might lead to responder confusion. For simple exploratory studies, Conjoint.ly recommends 2 to 5 attributes and 2 to 5 levels per attribute for an effective experimental design.
- Price does not need to be an attribute: You can use Conjoint.ly to test for any attribute in any type of setting, whether for consumer goods, services or for charities trying to find out donor preferences.
- Include an opt-out or “none of the above” option as the customer might not want to take any of the options in real-life. This is absolutely necessary for pricing. The only major exception is in medical circumstances where customers (or doctors) have to make a choice.
- More is not better: A moderate number of questions is best (around 10 but not more than 14). Having more than 14 questions causes responder fatigue and makes your results less useful. Conjoint.ly generates the optimal number of questions based on the setting you specify and also recommends a minimum number of respondents.
- Include some profiling questions to learn more about the demographics of your customers: e.g., age, income, education level, but don’t overdo it. Conjoint.ly lets you analyse these as well.
- Levels are like degrees of a characteristic and should be precise: e.g. the levels of the engine power of a car are 1.5L, 1.8L, 2.0L, not “less than 1.5L” or “more than 2L”.
- Levels need to be realistic and feasible to achieve real-life scenarios and should include competitive offerings.
- Try to avoid (or at least use sparingly) pairs of levels you want excluded from the survey: e.g. a car with an aubergine colour and sunroof must not be shown together as the company could never realistically supply this. Prohibitions limit the power of conjoint analysis. If you must include prohibited pairs, we recommend you use a branded study (be the first to know when we make branded studies available by signing up to updates)
- A picture is worth a thousand words. Use images if you have trouble describing your attributes. Conjoint.ly will soon allow you to do this easily (again, sign up for updates!)
- One of the great things about doing conjoint analysis is that it estimates market share based on customers’ preferences. Check out Conjoint.ly’s Excel market share simulator. But remember: conjoint-based market share simulation should not be relied on in isolation as it is nearly impossible to include every competitor’s offering in the market, and factors such as shelf availability of products do come in the way.
And, finally, when you’re doing conjoint analysis, always pre-test your survey with a handful of respondents (your colleagues) to ensure the survey fulfills the objectives of your investigation.
Read more Guides and how-tos.