Predictive analytics is a pretty amazing field of data science and has the potential to revolutionise the way we make decisions in business, so in this post I’d like to see if I can clarify just what the field of predictive analytics really is.
At its core…
Imagine for a moment you’re sitting at home planning your Christmas shopping. You’ve got a list of all the people you’d like to buy presents, and probably assigned each of them a rough dollar amount you’d like to spend on each. In your mind you’ve probably started working backwards from Christmas Day and figured out roughly what dates/times/places/websites you’d like to visit in order to find the perfect gift and then thought about how long it’ll take for those gifts to be found/packaged/wrapped/delivered so that they are ready for the big day. Without realising it, you have been performing many of the functions of predictive analytics.
Conceptually predictive analytics is precisely this: based on past experience, knowledge and insight, make predictions about a future event, including some kind of likelihood of this occurring.
It could be argued that businesses have been performing this function pretty much since the start of business. Each time a department store manager talks about getting ready for the holiday season, he or she is making a predictive statement based on past experience, knowledge and insight. I’m confident that the traders of Egypt and Babylon observed similar patterns in their own cultures and times and worked hard to maximise their returns based on this data. However, in the past 10 years or so, this kind of analysis has been put on steroids. These days its possible to not just predict trends accurately, but identify specific products to individuals, almost live. From a business perspective this is amazing! Simultaneously you can create efficiency, maximise sales and forecast far more accurately. There’s three components to this which I’ll expand on below.
It’s a matter of experience and knowledge…
The first component to predictive analytics is access to historical data. Put simply, the quantity, level of detail and accuracy of available data determines the accuracy of the predictions you might make. For instance, if you recorded every trip you made to a store for every day for 10 entire years, it’s pretty likely that you’d be able to predict quite accurately how much longer this trip might take over the Christmas period. Furthermore, you’d probably be able to make some observations about the general trend of these trips, the quality of the infrastructure and which set of traffic lights were the ‘worst’.
In a similar manner, the field of predictive analytics aims to relate previous data to future events. It does this through accessing as much data possible decision points or meaningful factors and then working out the trends, related outcomes and minute detail which the data indicates. This gives rise to the term ‘Big Data’ - a level and quantity of data which is so enormous it is literally beyond the human ability to understand, or put another way, the definition of ‘not being able to see the wood for the trees’. However the reason for this is pretty simple - the more data you have, the more accurate it is and longer it exists, the better results you will get.
Incidentally, when you read or hear about Amazon cloud technologies and the current race between Amazon Web Systems (https://aws.amazon.com), Microsofts Azure (https://azure.microsoft.com/en-us/?b=16.33) and Googles cloud technologies (https://cloud.google.com), you are hearing about the computing systems which make it possible to store, access and process these enormous data stores.
The second component to predictive analytics is the insight and it’s in this space that the true ‘magic’ of predictive analytics takes place. Insight is the ability to capture or make sense of the experience and knowledge gained previously. Returning to example of the trip to the shops, you gain insight when you take the knowledge and experience you’ve gained and apply it to your NEXT purchase. In a similar manner, the insight key of predictive analytics is the point where businesses can gain information to understand their choices.
A truly extensive analytics system can provide pretty amazing detail to businesses. Depending on the extent and range of your businesses you can start to understand all sorts of interesting things about what is happening behind the scenes of your organisation. This is especially true when elements of machine learning are introduced, as analytics systems systems can start to draw links between seemingly unrelated portions of businesses. In some reported examples, an analytics system was able to draw links between a supply chain issue and a particular factory producing substandard parts. In other cases, analytics systems has been used identify precise products to offer individuals, identify efficiencies in construction and whole range of interesting things.
In my mind, this is where predictive analytics is really heading. Assisting organisations in this field is pretty much my driving passion and I love the possibilities it opens up.
The final component to predictive analytics is what I tend to think of as the ‘useful’ part. Up until now, a predictive analytics system has been humming away, working through mountains of data and somehow making sense of it all. But in the predictive part two really important things happen.
Firstly, the system generates predictions. Put another way, this is where I get to see other people get excited about something that up until now they haven’t really understood. I’ll never forget the day I was able to show an engineering firm how to track a contract automatically, saving them millions of dollars in downtime and assisting their employees. The feel of excitement and relief was one of the highlights of my career so far!
Secondly, the system will generate some level of confidence for the predictions. This is a statistics term which recognises that the future is inherently uncertain and that there are ALWAYS outliers to a given range of results. There is always going to be that one customer who is unpredictable or changes his or her mind at the last minute. In most cases the level of confidence is set to such a level as to reduce any cost of incorrect predictions, however it is still something to keep in mind.
And so you have the basics of what predictive analytics is. I hope now you’re starting to see just how incredible this field and why it’s so interesting and exciting. I’ve got so much more to share on this topic so please feel free to drop me line or leave comments and I’ll respond when I can.
Some examples of predictive analytics at work…
Ubers redesign is really a UI function supporting predictive analytics
The cloud applications of Amazon, Azure and so on are really underpinning the massive amounts of data required to support predictive analytics