Derived importance uncovers the real drivers rather than telling us what the respondent feels we want to hear.
In other words, derived importance analysis goes beyond what respondents claim is important (i.e., stated importance) to them as it can uncover the underlying reasons for making consumer choices.
For instance, a respondent might state that ‘safety’ is very important in a car. However, this does not necessarily mean that ‘safety’ drives overall satisfaction or car purchases. This might be just a hygiene factor and actual purchases might be driven by other car elements.
Derived importance is assessed by measuring the relative impact of product features on an overall metric such as overall satisfaction, likelihood to recommend, or likelihood to purchase.
A statistical model is used to evaluate the relationship between a set of product features (the independent variables) and the overall metric (the dependent variable). For each attribute the statistical model provides a coefficient, which represents the relative importance associated to that attribute.
There are various statistical model that could be used to elicit derived importance scores, with the most popular being:
- Correlations: it is probably the most popular approach due to its simplicity. Unfortunately, it is also the most inaccurate as it explores the relationship of one attribute at-a-time against the overall metric. The inter-relationships are not assessed so the overlap contribution of two or more attributes on the overall metric is disregarded, even though this is often an important element, leading to unrealistic outcomes.
- Regression analysis: while this approach has the advantage that all attributes are included in the model, it provides extremely misleading and unstable results when the attributes are correlated to each other. Unfortunately, highly correlated attributes are present in virtually any survey (e.g., similar attributes or attributes exploring associated elements, such as quality/cost of a consumer product or safety/efficacy of a medical treatment).
- Shapley Values analysis: it is the most recommended approach as it provides robust and reliable results.
Derived importance analysis is probably one of the key areas in which the Consulting team has got long-standing experience. It routinely adopts advanced statistical models such as Shapley Values and Kano Analysis. The Consulting team can offer the package R-sw Drivers dedicated to derived importance analysis.