Data – particularly Big Data – isn’t that useful on its own. It needs to be analysed before it can be acted on, and we refer to the lessons that we learn from the analytics as insights. Insights inform the actions we take with the aim of creating business growth and driving efficiency.
As a basic example, consider a machine. Don’t worry about what the machine does – for the purposes of this discussion – all we need to know is that sometimes it breaks down. It’s fairly easy to collect one dataset showing the times that the machine fails, and another dataset showing what the machine is doing. By analysing both datasets, we can see what activities are likely to make the machine break down and take action to prevent it from happening.
Data scientists refer to the analytics involved in this simple case as “descriptive analytics”. It describes to us what has happened, when it has happened, perhaps even how it has happened or why it has happened. What it doesn’t do, though, is tell us directly what we need to do to reduce the chance of it happening.
Most of the analytics activity carried out by businesses and organisations today is descriptive. At times we may feel like we’re swamped by reports and charts showing us how often a particular event (for example a sale, or a customer social interaction, or a mechanical failure) took place. That’s as far as it goes – and it’s down to our human brains to work out the rest.
Descriptive analytics can certainly be very useful, particularly if it is used in a strategic manner. Google Analytics is a good example that a lot of people will be familiar with – it tells us who is visiting our websites, what times they are most busy, and what visitors did while they were there.
But as organisations become more digitally mature and able to deploy more advanced analytics technology, it becomes possible to look forward to what is likely to happen in the future.
This brings us to “predictive analytics” – often the goal of organisations moving into the sphere of artificial intelligence and machine learning.
The basic premise of predictive analytics is that it takes data we do have – information about what has happened – and extrapolates from it to fill in data we don’t have. These are predictions – “best guesses” about what is likely to happen in the future.
So, taking the example of the unreliable machine we started out with, we have data for how and when the machine failed in January, February and March. After running it through a predictive algorithm, we end up with data on when and how it is likely to fail in April, May and June.
(It’s important to bear in mind that predictive analytics is always probabilistic. Of course, it can’t tell us with 100% certainty exactly what will happen. What it can tell us is, based on past performance, what is likely to happen - usually with a stated degree of probability).
This is very helpful of course – if we know when our machine is likely to break down we can ensure we have spare parts in place to repair it, and contingencies to let us carry on working while it’s being fixed.
Another everyday example of predictive analytics in action are the credit-scoring mechanisms used by banks and lenders to assess the risk of people applying for credit. The analytics provide an estimate of the applicant’s likelihood of making their repayments, and the bank decides whether or not it exceeds their risk threshold.
Although it’s becoming more common to see this sort of analytics technology being put to use, the truth is that until very recently only the most well-resourced and skilled businesses have been advanced enough to implement and see value from it.
Today, however, the most forward-thinking organisations are taking a step further still, into the realm of what is known as “prescriptive analytics”.
So, predictive analytics tells us what’s likely to happen – but it doesn’t tell us what the best course of action is to achieve an optimal outcome.
The next step on the analytics maturity ladder does just that. While a predictive analytics system will give us a range of possible outcomes, it doesn’t know which is the best one to take. Sometimes this is fine, because the people receiving the insights will know what to do next. If the aim is to increase sales volume, they can choose the action that produces the highest volume of sales. If the aim is to increase the average value of each individual sale, they can choose the action that increases that.
But if the aim is broader – such as to increase overall revenue – and they don’t know whether it’s best to do that through increasing sales volume or sales value, a prescriptive analytics solution would be the way forward.
Autonomous vehicles are a good example of analytics-based systems where prescriptive analytics are essential. It isn’t enough for the car to “know” that turning left at a junction is the quickest direct route to a destination, but also that it runs the highest risk of encountering heavy traffic and lengthening the journey. It has to be able to choose the best course of action and “prescribe” it to the computer controlling the car’s movement, without the need for human intervention.
Returning for our last time to the example of the accident-prone machine, when kitted out with prescriptive analytics, the person operating it (or the machine itself, if fully autonomous) now knows not just what has caused it to break down in the past, and what is likely to cause it to break down in the future, but what the best course of action is to keep the delay and disruption caused by future breakdowns to a minimum.
Those are three basic levels of analytics maturity that organisations may go through, on their road to automation. Until recently, the more advanced predictive and prescriptive techniques may have been prohibitively expensive for all but the biggest companies. But tools and platforms are increasingly coming onto the market that mean smaller operations can at least get a taste of their usefulness. If you use social media for marketing purposes, then there are tools that will analyse your followers and activity to suggest the best content to post, and the best times to post it. And if your business relies on digital transactions, tools exist that proactively warn against, or block, transfers that display patterns that suggest they could be fraudulent.
Moving step-by-step up the ladder from descriptive, to predictive and eventually to prescriptive analytics is an important part of the journey towards managing a successful digital transformation.
Bernard Marr is an internationally bestselling author, futurist, keynote speaker, and strategic advisor to companies and governments. He advises and coaches many of the world’s best-known organisations on strategy, digital transformation and business performance. LinkedIn has recently ranked Bernard as one of the top 5 business influencers in the world and the No 1 influencer in the UK. He has authored 16 best-selling books, is a frequent contributor to the World Economic Forum and writes a regular column for Forbes. Every day Bernard actively engages his almost 2 million social media followers and shares content that reaches millions of readers.