What About the Little Result?

Jun 19, 2012 by David Loshin in Data Management, Data Quality

In the last post, I posed that the high-level, aggregate values used for performance measures (what I called “big results”) are often easier to get than the component parts that roll up into those measures (which I called “little results”). The interpolation of the little result from the big result might open up a wide range of opportunity for creating value.

As an example, let me use one of the sample restaurant measures I used last time: average check per guest. This little result could be a function of two big results: the total amount of restaurant food sales (accumulated from the register totals by store) and the number of guests (which could be counted using an electronic counter whenever someone enters the restaurant, or could be simply counted by each store’s manager). Divide the total restaurant sales by the number of guests and you have the average check per guest.

However, this does not tell you much about your guests. Some of them may be good regular customers, coming in once or twice a week with other guests and buying many items. Others may be less good – individuals coming in, just ordering coffee, and occupying a booth for 3-4 hours at a time. Clearly, some types of customers are more desirable than others. But in the aggregate average check per person, those differences are completely washed out.

In other words, even though the big result is pretty trustworthy, the interpolated little result is skewed away from reality. Once customer spends $10, the other spends $1, the average check is $5.50, which is not close to either. But does the difference really matter? We’ll examine that in the next post.

Tags: Data Management, Data Quality