Novelty, Utility, Insight, and Interest.

There are several reasons to look for patterns in data and modelling it for business. One could be clarifying the business performance, another to determine opportunities for improvement, yet another could be to determine emerging problems of control.

The expected insight is never ‘keep doing more of the same’. The real question is often ‘what else we have not thought of that we should do?’ Novelty is in the eyes of the beholder. Utility is equally difficult to determine. Insight and interest are susceptible to estimation.

Data science discoveries that are interesting and insightful—usually carry more information that those that are not. Data science or mining is not mere search for patterns. The patterns must be placed in perspective within the model. For example—the business problem could just be ‘customers are complaining because it takes longer to settle their claim’.

Yet it could be hard to exactly find what is going wrong, where, and at what part of the claims process?

So, we need to construct the frame before we attempt to model. There are several ways to go about this—the systems, open and closed forms of systems, static and dynamic patterns, coupling and feedback—however this construct calls for much detail and we will not cover it on this blog.

Just to clarify the concept of a system—let us resume our premia / insurance sales example.

In the figure-sales sit at the centre and driving factors are around the outside. Does this work?

Think about product quality. If the product quality is high, Your firm can sale more.

Conversely if Your firm can sell more, they can maintain the high quality.

That too drives a motivated sales force. These are all interlinked loops.

What is needed here is some way of looking at this whole relationship—which is an important skill for the data scientist to develop.

For fun, do you want to draw similar pictures for the part you play at Your firm?