Syndicated from Smart Data Collective
I often hear people talk about analytics, especially advanced analytics like data mining or predictive analytic modeling, as though the value comes from “aha moments”. Sudden moments of clarity, defining moments, where the analytics deliver some piece of dramatic insight that enables a company to see some fantastic new market opportunity or fundamentally change the way it does something. This is a myth. Most companies that get value from analytics either do so without ever having an aha moment or they have both an aha moment and something more – the aha moment is not enough on its own.
Let’s take the first of these circumstances. Many of the companies I work with (as well as many that I read about) are getting tremendous value from analytics. But they are getting this value by applying analytics to improving decisions, operational decisions, that they make in huge numbers. They use analytics to better predict the fraud risk of a claim so they can pay fewer fraudulent claims. They use analytics to better predict credit risk so they can manage credit lines more effectively. They use analytics to predict which cross-sell offer will be most likely to succeed. There’s no great aha moment, no one thing that analytics teach them about their business. What there is, instead, is an industrialized approach to analytics that focuses on building models that will predict something useful about a single customer or single transaction. There is an implementation process to make sure these predictions are turned into useful actions, often by applying business rules, and there is a feedback or test-and-learn process to constantly evaluate how well the model and the actions based on it are working.
In the second kind of company there is an aha moment – analytics tell them that they have far more fraud than they expected or that a particular product is not as profitable for certain customers as they believed. But this aha moment is not actionable without understanding the operational decisions that must be made differently and without exactly the kind of process I just described to make sure those operational decisions are managed. What both kinds of companies have in common is a focus on operational decisions as a source of competitive advantage.
Now, obviously, I am being provocative. Some companies do, in fact, get value from analytics from a single aha moment. There are drug companies that find the group of patients that a drug will work for, insurance companies that find that one particular policy type is causing huge losses, medical professionals that find the one thing that predicts a negative outcome to a treatment and so on. Most do not. So when you think about analytics, don’t think about aha moments, think about the operational, transactional, micro-decisions that drive your front-line systems and think about how analytics could make each of those decisions a little better.
Comments on this entry are closed.
James, human decisions aren’t done with logic. They are done with emotional intuition. (I can give you a lot of literature on that.) That is not a defect, but evolution found that it is the most effective way to go through life. The world is built on uncertainty, unpredictability and emerging structures that simple Booelan Logic or statistical observation can not deal with. It never could.
So the statistically driven decisions not to give someone a loan based on a credit score is not a better decision, it is simply one that will fit better into future statistical analysis. That means we are talking about self-fulfilling prophecies here. We are also delegating our responsibility to be human to a computer and that is the worst part of it. We are just scared s#!tless to make a wrong decision and stand by it. So we delegate to a computer and then it is not our fault.
I wonder how long humanity will survive that nonsense …
If you think human decisions are so great I suggest you go talk to a person of color or someone with an unusual career or lifestyle (or even dress sense) and see how good they think most people when they use intuition to make decisions. Only some people can use their intuition effectively, most people use “what everyone knows” kinds of approaches and these are biased, bigoted, inaccurate and poorly matched to objective reality. A statistically based loan decision is a better decision from the point of view of the loan issuer if they get better business results out of it.
Oh , and one more thing, the recent crisis was largely caused by people whose intuition said that house prices would continue to rise (because they wanted it to, another problem with intuition) and so overrode what the models were telling them – they used their intuition to change the models to get the answers they wanted. Not really a problem with models so much as with humans.