What is an Insight? Going Beyond Descriptive Data and Analytics
Generate Insight by Framing an Analytic Problem to Include Actions
As a marketing analytics consultant, I have heard numerous times the problem of analytics. Many CMOs / CEOs have said:
“Most of what I get is too little too late.”
“The analysis is not really relevant to my business.”
“The analytics have no action-ability.”
“It’s too complicated / sophisticated and buried in statistical minutia.”
“Gives no plan of attack, just presents problems.”
“Mostly data presented with no cause or explanation.”
And of course they’re right. I have also heard clients report the following insights:
“92% of our customers wear some kind of jeans when they come into our store.”
“Net revenue decreased over 3.5% YOY in same store sales.”
“We have the highest rated product in our industry!”
“Our customers trust our brand more than all other brands.”
“Both market research and independent field tests show treatment X outperforms treatment Y.”
Note that all of the above insights are really just observations. This is the problem of insights--they are confused with mere observations.
There are three general uses of data and analysis: descriptive, predictive and prescriptive. These each have a place and the distinction between them needs to be clear and obvious.
Descriptive data and analysis is looking only at past information. It is where Business Intelligence and reporting live. It is a necessary but not sufficient part of insight generation.
To be an insight there has to be element BEYOND descriptive data and analysis. That is, an insight has to be, at least, predictive. Predictive analysis is what my book Marketing Analytics is all about. To be predictive there needs to be a quantification of causality. To be predictive and insight-generating there needs to be independent variables that we can pull to change the estimated (predictive) dependent variable. In terms of say ordinary regression there needs to be, for example, price as an independent variable that CAUSES the movement in the dependent variable of UNITS. Note that price is a lever under marketing’s control to move, giving them an action.
Prescriptive data and analysis is about optimizing some over-all metric, typically profit margin, etc. It is usually done via a system of equations, simultaneous, structural or otherwise. It attempts to simulate the system and turn out optimizing levels of factors in order to maximize some KPI.
So, in order to be an insight the analysis has to be either predictive and / or prescriptive. Otherwise it’s probably just an observation.
I would offer the below as a guiding definition of what is an insight:
- Actionable definition: Something new that gives actionable ideas. It has to be both new and provide a call to action.
- Understanding definition: Generates a fuller description of behavior and fleshes out a (consumer) decision process. Note the hypothesis of causality assumed here.
- Overall definition: Insights provide new information / understanding by explaining consumer behavior and quantifying causality with actions that drive a strategic advantage.
In order for analytics to be used, there has to be insight-generation. This means the analysis has to be at least predictive. This means the framing of the analytic problem has to include actions. This means the analysis has to quantify (typically consumer behavior) causality. When you present analysis to a senior business leader and are talking more about actions they can take to change the direction of their business instead of significant t-ratios and adjusted R2 metrics, you are providing true insights.