Editor’s note: Zoe Beales is research manager at Northstar Research, London.
I attended a three-day training session in Rotterdam, Netherlands (put on by SKIM and Sawtooth), all about conjoint and choice-based methodologies. Attendees came from a variety of backgrounds, including consumer, pharmaceutical and academic with mixed experience in the technique and Sawtooth software.
If you’re new to conjoint, here are seven tips based on what I learned:
1. Conjoint has many uses. Conjoint is often associated with consumer research and asking participants to select which product they would buy in a given scenario. However, it can be used across multiple industries and does not have to be purely focused on buying products. Other examples of usage are: designing benefits packages for employees and developing new pharmaceutical treatments.
2. Remember sampling basics. With advanced methods like conjoint, don’t forget sampling basics! You don’t want participants to be selecting “none” on every task because the product you are testing isn’t relevant for them. Make sure you are targeting the correct participants by screening appropriately in your survey.
3. Focus on objectives. Making sure your business objectives are at the center of your design was constantly repeated. Although loose rules exist in deciding on number of attributes, sample size and number of tasks, do not lose sight of the main purpose of your research and ensure you take an approach focused on business objectives.
4. Don’t overload on content. Conjoints are made up of multiple attributes which are essentially the different categories of a product (e.g., brand, color, size, etc.) and within each of these are levels of the different options. A temptation from internal stakeholders is to try and include as many attributes as possible in a conjoint but this can have a negative impact on the results. Remember, the business objectives need to be at the center of the design so prioritize attributes that are central to these. If there are additional attributes which are not central to the main objectives, they can potentially be examined in a separate survey question.
5. Use prohibitions sparingly. As always, this piece of advice will depend on your business objectives but generally try to limit use of prohibitions (this is excluding certain combinations of levels being shown together). Ideally you want to limit any correlation between levels as this will have an impact on your results. Providing the combinations of levels make sense and could exist then the results can provide insight. To reassure your internal stakeholders if they are concerned they would not introduce a product with certain levels, you would just not include in the market simulation (the main analysis tool used for conjoint which allows you to simulate scenarios).
6. Remember survey design fundamentals. The standard rules of survey design still apply! As with any survey, if it is too long and does not make sense then you risk your participants being unengaged and potentially dropping out. Before launching any conjoint work with someone not involved in developing the attributes and levels to make sure it makes sense. If the conjoint is part of a larger survey it is also important to ensure engagement does not decrease before getting to the conjoint.
7. Explain outputs clearly. As well as the market simulator, other outputs from conjoint include:
Utility scores – within each attribute these scores allow you to measure preference. The scores are interval data, so just like when discussing temperature you cannot say a score of 20 is twice as good as 10.
Importance scores – a calculation using the utility score to work out relative importance of attributes.
Both scores can be hard to interpret so be cautious when sharing with less research-savvy internal stakeholders. Even if you have provided a clear explanation to your immediate internal stakeholders, the scores can then be shared across the business without this explanation and therefore create confusion! Instead, encourage internal stakeholders to focus analysis around the market simulator.