Last week, we automatically identified similarities over a period of time with the help of a computer. Today, we will define an indicator that will help us analyze differences.
Although we could analyze differences using numerical values such as revenues, margin, costs or quantities, we can also use subjective rankings, for example, those obtained from a customer questionnaire. Let’s assume that we have conducted a survey in which customers have ranked their satisfaction with Product X on a scale of 1 (excellent) to 6 (poor). Each of these customers can be described using a multitude of attributes including age, income, job, gender and location.
Our objective is to learn if the satisfaction level of a given customer is somehow related to his or her attributes. Although the typical approach is to create a hypothesis and then test it using various statistical methods, we can learn more by using a simple ranking. (Our smart indicator will do the hard work for us.)
The mean customer satisfaction is 2.3, but varies significantly in rural regions (1.2) and urban areas (3.4). This type of benchmark calculated for various attributes and also for combinations of attributes (e.g. “big city” + “average income”) may let us find many interesting differences in our customer base.
First, we need to calculate the variance between the mean customer satisfaction and each possible combination of customer attributes. This multidimensional task can be easily executed in an OLAP database using a calculated KPI – variance from mean. Using this KPI, we can use a simple ranking to determine which customers are more and less satisfied than the mean.
Best of all, the results are easy to understand, because the complexity of the question is packed into the indicator – and not into the statistical analysis.