Binomial distribution for a KPI status (parts 1 and 2)

This entry is not much about OLAP or MDX but gives some background on statistical tests that may be very useful in designing meaningful KPIs based on statistical hypothesis tests. In KPIs we usually compare actuals with goals and visualize the results as a traffic light. In most cases both the actual value and the target can be read from the cube. It is very common to have a dimension "Scenario" with members Actuals and Forecast or something like that.

But when dealing with statistical results, things are getting a little bit more difficult. For example, we're testing the quality of products in a manufacturing line. If the quality test is expensive we have to rely on a spot sample. Let's assume we want a quality of 99%. This means that only 1 of 100 produced items should fail the quality test. So for our spot test let's say we pick 500 out of 10000 produced items of a charge and perform the quality check on them. Depending on the result of the test we decide if we are going to keep the remaining 9500 items or not.

Binomial distribution for a KPI status - part 2