Who doesn’t want to be proactive?
Many a starry-eyed presenter have discussed the subject of predictive decision making. Usually, the story goes something like this: “You can’t drive your car looking in the rear-view mirror. You have to look out the wind screen. Therefore, you need predictive decision making.” Examples of how this is actually achieved are then omitted, or breezed over, as the value of the approach should be obvious to everybody. Indeed, who would want to drive using the rear-view mirror? Who doesn’t want to be proactive in their decision making?
It’s not as easy as it seems…
The problem? Once they start examining how to go ahead in reality, most people quickly run into problems. It doesn’t look to be as easy or obvious as it seemed when the car analogy was made. And indeed, there is a difference between a company and a car. Well, many differences really, but one that stands out in this case is the company’s lack of a wind screen.
Because: The data we collect are really showing what has already happened! Any attempt to look into the future is actually using that data to extrapolate events that are to come. In other words: We are making a forecast or predictive model. And, traditionally, making a prognosis requires some level of statistical proficiency to make it more than wild guesswork. Which is where it stops being simple.
…Or does it? Does analytics really have to be that complicated?
I would argue no. The only option is not to hire more statisticians. There are many user-friendly tools on the market today that will allow users with a good understanding of the business to employ the power of predictive (and descriptive) analytics and advanced self-service. Many of them also have the need to understand how future events are likely to be affected by decisions made today.
It’s basically all the people that are building spreadsheets, constructing their own access databases, or repeatedly sending out ad-hoc requests for data or new reports. There are many candidates:
- Marketing people
- Product and service owners
- Sales managers
- …and the list goes on
A shift in the framework for decision support
What’s needed is to start looking beyond the capabilities of simple dashboarding and ad-hoc reporting for these people. Companies need to understand the difference between managing their everyday processes through various KPI’s and performance management initiatives, and supplying their decision makers with the information they need at the point where their decisions are made.
The first requires a solid and stable framework that is reasonably (if by no means perfectly) well suited for traditional IT waterfall development models. In this scenario, IT supplies a set of reports based on the company’s steering model.
The latter requires a new approach to business analytics, where flexibility and agility should be prioritized over structured, and where ingrained models for report development need to be challenged. Here IT supplies data and the capabilities needed to get whatever value out of that data that the present situation requires.
It’s this second scenario that should be explored by organizations. And this is where the predictive capabilities for the statistically challenged should reside. And in fact, the same tools can add great value to the ‘real’ analysts too!
The predictive laboratory
It may seem like an impossible equation to make these two realities coexist, but the view on that is shifting. The concept of the analytics laboratory – or sandbox – is becoming more and more frequent in discussions in many forums. Basically, it consists of a separate window into the data, living in close relation to the existing information delivery mechanisms. Specifically, existing data warehouse and information management investments should be leveraged to the greatest extent possible in the laboratory too, but when that is not enough (which is surprisingly often in many organizations), a greater degree of freedom is allowed to get quick results. Because, at the end of the day, that is what is going to make a difference.