In recent months we’ve been seeing a rise in the amount of interest in predictive analytics from our clients, so we’ve decided to run a series of blog postings to explain what predictive analytics is, how it’s used and so on. This posting is the fourth in the series.  In this post we’re looking briefly at how several future-oriented techniques differ.

 

First of all strategic planning, the process an organisation goes through to set its mission, objectives and strategy. Strategic planning is based on establishing where you want to go, and so is based on your will and intent.  It covers a very broad set of future events.  It may be that there is no historical data to base these decisions on, but something novel has happened and a committed group of entrepreneurs decide to commit to a goal, based on a shared vision of how things should be done.

In contrast, predictive analytics is rooted in fact.  It usually involves answering a tightly defined question, a “target”.  It is based on data points of events that have already happened.  It applies statistical techniques to predict the probability that a desired event will happen, or to predict the most likely outcome whether it is desired or not.

As for forecasting, it depends what you mean by forecasting.  For many people in businesses, forecasting is often tightly coupled with the word budgeting.
Usually budgetary questions are quite specific, but broader than those tackled using predictive analysis.  Often coarse techniques including manual intervention are used to create financial budgets, and in many cases they are not linked with historical data points at all.  When used in the sense of
budgeting, forecasting lies somewhere between the two extremes of strategic planning and predictive analysis.

So what?  Let me give an example of how predictive analytics techniques is changing things.  Some time ago I was involved in developing a business intelligence system for managing the delivery of activity against a number of contracts.  The client was responsible for around 50,000 contracts and the terms of each contract were set within a framework but were themselves unique.  The client wanted to be notified on a dashboard if any term on any contract was broken, and that requirement was successfully met.  The client also wanted to be notified if “something out of the ordinary” happened on any contract, and this requirement proved much more difficult to satisfy because there was no straightforward way of defining “out of the ordinary”, other than by reference to events that had already happened in that specific contract.

Using the techniques at the heart of predictive analytics, in some circumstances it is possible to handle this latter requirement too.  If enough meaningful historical data is available, a specific model can automatically be constructed for each contract and new data checked for deviation from that particular pattern of behaviour.  Previously, constructing individual models took a long time, but recently it has become possible to automate the process of creating optimised models.  This has greatly reduced the time and effort involved.

In other words, predictive analytics is now enabling the move from general comparison against a whole population, to personalised comparison at individual level.  This opens up possibilities like budgetary target values to be set at a much lower level of granularity because the productivity of people in the financial planning team is greatly increased.  It could also lead to more informed strategic planning decisions being made because the out-workings of scenarios can be seen much more quickly through automated modelling.

Next time we will begin a number of postings looking in more detail at some of the techniques involved in predictive analytics.  The first will be tackling the question, “What are the main steps involved in carrying out predictive analytics?”