In a previous post we examined the fundamentals of Likert scale development. In case you missed it, Likert survey questions are designed to measure attitudes using a five or seven-point scale agreement scale (e.g. strongly disagree to strongly agree). These scales and their derivatives continue to be used extensively in both consumer and B2B market research.
As the old song goes, attitudes lead to intentions which lead to behaviors, which can lead back to attitudes. With the advent of customer relationship management systems (CRM) and marketing data warehouses, market researchers can compare stated attitudes to behaviors or transactions. This allows us to ask such questions as:
- Do higher levels of customer satisfaction lead to greater profitability?
- Is this consistent across all market segments?
- What is the average time to second purchase for a new customer?
- Is this impacted by their attitudes toward their first purchase experience?
Our last post on the topic ended with thoughts around further development and refinement of attitudinal scales. Given that survey real estate is at a premium, we want to develop a scale that is both valid (it measures what we think it is measuring) and reliable (it measures consistently over repeated applications). Only when both criteria are met should we fully implement our scale.
Pre-testing items for a scale allows the researcher to narrow the focus and keep only those items that have the highest level of input. This can be accomplished using Cronbach’s Alpha, which provides a measure of the scale’s reliability. Output from this test (which is common in most statistical packages) highlights those items which could be deleted without loss of scale reliability.
Another test to consider is principal components analysis (PCA). This is a closely related cousin to factor analysis. If we include a large number of items in our initial test then PCA can be used to highlight the underlying structure. Our group of items may in fact be measuring a smaller set of factors. Output from this test allows us to identify those factors and which items are most closely related to each factor (or dimension). A full treatment is beyond the scope of this post, but more information can be found at Wikipedia.
We want to ensure we are measuring those attitudes most closely related to a key criterion, e.g. loyalty, satisfaction, profitability, likelihood to recommend, etc. This process requires testing and calibration, but ultimately will lead to survey data that has an impact on the bottom line.