For analyzing, understanding and operating any business, there are three critical “grains” of interest - time, entities, and the all important grain of the underlying business process itself.
Let’s dive into these concepts:
I. Time grains are widely understood. The utility for varying time granularity is driven by:
a) The business operating cadence: as an example, weekly or monthly time grains are used during periodic business reviews
b) Business initiatives or constraints: it is common to compare time windows based on campaigns or feature releases, or to drill into specific time ranges that operated under different business constraints or conditions.
c) The nature of the process: as an example, you may need to control for expected variations for highly seasonal processes like say fashion e-commerce.
II. Entity grains are better understood as “segmentation” - that is, via the grouping of a set of entities. There are at least two variants to consider here:
a) Static segmentation:
Entity demographic or profile information, or attributes that can be permanently tagged like marketing channels are examples of “static” segmentation.
b) Behavioral and dynamic segmentation:
Grouping entities based on their behaviors can be really powerful. As an example, you can cluster high and low LTV users to find patterns and actionable insights.
III. Process grains:
Businesses are complex inter-connected systems but a specific analysis often zooms into a “node” in the overall networked process, which is often the atomic process grain. There are at least two concepts at play when examining process grains:
a) Direct relationships: processes like an acquisition flow where entities move through a series of steps are examples of clear direct relationships across atomic process grains.
b) Indirect relationships: processes like responsive customer support that indirectly relate to the retention metrics are examples where the exact relationship between atomic process grains is unknown.
Over the last 10 years, with the ease of cloud storage and compute, and the relative democratization of data querying skills, the ability to analyze across varying time grains or even entity grains via segmentation has become significantly easier.
The next frontier is making it effortless to traverse the process grains, to zoom in and out of the networked, inter-connected business processes.
