5. How To use « TIMi – StarDust module »? > 5.5. How to create a segmentation? > 5.5.7. Explaining the segments from a business-point-of-view

You can explain easily the segments using the intuitive charts explained in sections 5.2.10, 5.2.11 and 5.2.12. All these charts can directly be added to the final report in 1 mouse-click. Most of these charts can directly be exported to Excel in one mouse click.

A good “trick” to setup rapidly all the charts is to convert the segments into groups, as explained in section 5.3.1.5.

The charts of section 5.2.10, 5.2.11 and 5.2.12 allow you to see some of the best explanatory variables for each segment. You can use these charts to select the variables to “present” to the business-users to “explain” in the simplest way your segmentation. These charts are good for a preliminary analysis of your segments but they have some limitations:

1.You don’t know how many variables to “present” to the business-users to “completely describe” a specific segment.

2.When you are “describing” a specific segment, you should avoid to present inside your report two correlated variables because they are describing the same “concept” related to your segment. Correlated variables will both have approximately the same (high) KL value and, thus, you will be “tempted” to include both variables inside you report. Thus, if you are using only the charts from sections 5.2.10, 5.2.11 and 5.2.12, you will have to do arbitrary and subjective choices about the variables to “present” to correctly “describe” a specific segment. There exists an objective and automated way of selecting the “right” set of variables to “describe” a specific segment. This “optimal” set of variables contains the smallest amount possible of different variables that characterizes a specific segment, the “best as possible”: see section 5.5.5.1 to obtain this “optimal” subset of variables.

You can “note” all your findings and “insights about your customers” (to be sure to forget nothing) inside the small HTML notepad included with StarDust (see section 5.2.12).

You can switch rapidly between different segmentations using the technique described in section 5.5.2.4 (Control 4: Random seed).