A recent post from Thomas Anderson on Analytic Bridge blog. Thought this was quite interesting and wanted to share with my readers.
Many BI platform and analytics vendors talk about the ways in which their tools enable people to acquire insights by means of analytics. That sounds like a good thing, but what do they actually mean by insights and why are insights associated with such high business value? I’m curious to know what people think about this and whether there exist any established models for measurement of insights and/or their value. Some thoughts of my own are found below.
Assuming that analytics is about making business decisions based on data (as opposed to making business decisions that are not based on data), it seems to me that the concept of insight has at least three distinct meanings – each associated with a different degree of value – in the analytics space.
In what might be the simplest case, insight refers to discovery of relevant business information. For example, you use an analytic environment of some sort (report, dashboard or analytic application) to look at quarterly aggregated sales data by region and product and discover that sales of product A in region West is ten times higher than sales of product B in region East. Your own knowledge of the business tells you that this is interesting, surprising or relevant in some other sense. However, such an insight does not in itself tell you why a certain business situation exists or what to do about it. Hence the value of such an insight seems limited although it may be a starting point for more analytics that might generate more valuable insights.
In what might be the most common case, insight refers to discovery of a cause-effect relationship associated with some degree of reliability. For example, the analytic environment and your own knowledge of the business tells you that sales is linked to sales force size, time of year and product marketing budget in a certain way. You might be able to capture such an insight, e.g., as a predictive model, and extend the analytic environment with more or less reliable business knowledge beyond that of your own. The model may not be perfect, but to some degree it still allows you to explain what has happened and predict what will happen. Hence the value of such an insight seems fairly high even if it does not tell you what to do about a certain business situation.
In what might be the most interesting case, insight refers to a thoroughly substantiated and well formatted message to decision makers regarding a certain business situation. For example, the analytic environment, your own knowledge of the business, and other sources of more or less reliable business knowledge, e.g., predictive models, tells you that there are three plausible ways to increase overall sales, each associated with some degree of probable payoff and some degree of risk. Furthermore, the analytic environment enables you to present the insight to decision makers in such a way that the organization as a whole can share its nature, take action and measure its value. Such an insight seems “complete” and associated with very high value.
Does anyone agree or disagree with the distinctions between these different types of insights? Would it be appropriate to say that what essentially makes a difference between the first two is the ability to capture and reuse business knowledge within the analytic environment, and that what essentially makes a difference between the last two is the ability of the organization as a whole to embrace an analytic culture? Or is there an altogether different approach that makes more sense?