Similar to the age-old question of how many angels can dance on the head of a pin, in the world of analytics, the answer to the question of how many metrics are necessary for a given scenario can be surprisingly daunting.
Let’s start simply, and assume the scenario is an order process wherein any given order is initially “opened,” then continues to “ready for filling,” and goes on to “shipped.” Now, someone wants “counts” for the order status values. What counts? The person could be asking for a count for a given day, week, month … of how many orders were placed into an “opened” status, a “ready for filling” status, and a “shipped” status. Alternatively, that same person could be asking for, at a given point in time, the number of orders in each state, which is quite a different number. One set is a count of actions happening, the other is an inventory of created order objects. A single order going through multiple states in a single day would count more than once in the former, but only once in the latter.
Others may want to know how long an order is in each status before moving on to the next. One order, three status values, and we already have possibilities of at least nine differing metrics that may be useful. If we add in the possibility of an order expanding status values—for example, “canceled”—there are more combinations to consider. And then, add in a possibility of an “opened” order being able to jump back and forth into a “hold” status, or even a “ready for filling” to be on “hold” for backordered merchandise, and there can be an explosion of dozens of metrics that could be of value for analysis.
Calculating an “average order completion time” may be quite complicated. Such a calculation may involve using the date and time of the order being “shipped” and subtracting the “datetime” for when the order was “opened.” The derived gross duration of the order may still need some alterations; the cumulative time an order may have been on “hold” status might need to be removed from the gross duration to arrive at an adjusted order duration value. Variations may involve starting the duration metric at the point the order moves into the “ready for filling” status, should one be interested in the efficiency of the fulfillment process, versus the entirety of the order lifecycle. Each segment of an order lifecycle may be analyzed independently or linked in various combinations to be used in probing metrics for specific workflow paths. Variety is the spice of life, and it is the spice of analytics too.
Subtle Differences Abound
A good data modeler must know when there is something the users need to analyze, even when what that is is not necessarily obvious. Subtle differences abound between things in a state, versus actions taken to create or change states, versus the duration of a business object within or across states, versus business objects and the workflows containing them. Each of these subtleties drives differing metrics with distinct uses.
Beyond the described simple order scenario, if we dive down into metrics for each item ordered, quantities of items ordered, amounts for purchase or wholesale, these additional piece-parts could easily double/triple/quadruple our metrics, forcing us to enter a new transfinite metric reality. Quite a few metrics will be needed to define, track, and calculate important aspects of the complete order lifecycle for an organization. The most straightforward and simple of process flows can easily generate a very large number of metrics that have value to an organization. The designer wishing to provide a database home for standardizing metrics must embrace variety and always find a place for necessary measures.