In an era when data is seen as an increasingly valuable commodity, the tools and techniques used to analyse that data gain extra importance.
Historically analysis, even using computers, has been something of a labour intensive task because the raw data needed to be verified and complex models built to process it. But due to the growth of machine learning and data mining techniques we’re now seeing an increase in predictive analysis where machines can take historical and current information and apply it to a model to predict future trends.
Whilst in business terms predictive analytics is primarily about looking to the future – spotting sales trends for example – it has other uses too. It can be used to spot trends that might indicate fraudulent credit card use for example. As with any other type of data processing the quality of the results depend heavily on the integrity of the data and the quality of the underlying model.
Predictive analysis relies on the use of models to ‘score’ data. These models come in three main types. The most common is the predictive model which uses sample data with known attributes to train the model so that it can then look at new data and determine how it’s likely to behave. This information can be used to predict customer behaviour or spot patterns of activity.
Descriptive models are used to classify data into groups. Where predictive models focus on the behaviour of an individual customer, to determine credit ratings for example, descriptive models sort customers by characteristics like their age or their past buying behaviour. This information can then be used in marketing campaigns aimed at specific groupings.
Decision models take a wider view, taking account of the known data as well as results from predictive models. This allows them to forecast the results of decisions that involve a large number of different variables.
Within these models various statistical techniques are at the heart of processing the data. The regression model is the most common, used to define the relationship between variables. In addition time series modelling is used to forecast the way variables will behave in future.
Predictive analysis is closely tied to other technologies, in particular machine learning (opens in new tab). Data mining too is an important part of unlocking the value contained in business systems. As more powerful systems become readily available technologies like neural networks will be applied to build more complex analysis models.
Commercial analytics tools are available from many of the big IT suppliers. These include SAP HANA, Oracle Advanced Analytics and IBM SPSS Modeler. Open source tools are also on offer from developers including GNU and Apache.
Predictive analytics has been used in many areas of business in recent years. It’s frequently found in customer relationship management (opens in new tab) (CRM) systems where it can help in building sales campaigns, or providing customer services. Analytics can spot buying patterns and identify problems that could lead to lost business.
The internet has made it much easier for customers to compare services and products from different suppliers, so increasingly companies are looking at how they can retain existing business. Predictive systems can help here too, by looking at buying records and service use patterns it’s possible to spot when someone is perhaps thinking about switching suppliers. They can then be targeted with marketing initiatives to persuade them to stay.
In some businesses predictive analysis is aimed more at preventing problems than at gaining new business. A 2014 survey by SAP (opens in new tab) revealed that just over 40 percent of financial services businesses used predictive analytics to minimise risk. We’ve already touched on the use of predictive systems to spot fraudulent activity, the US Internal Revenue Service is known to use predictive analysis on tax returns to spot suspicious activity, but they have other uses in this sector too. Predictive systems can identify those customers who are likely to make late payments for example and help focus collection efforts. For insurance providers the technology can be used in the identification of risk and thus help to calculate premiums more accurately.
The medical sector is also starting to use predictive analysis techniques. On a macro scale they can be used to predict what proportion of the population is likely to develop certain conditions such as heart disease or diabetes. But they can also support clinical decision making, using patient data to plan long-term treatment.
Another field where predictive analytics has the scope to make a big difference is project management. Here it can be used to predict the risks of project failures, overspends and so on, data which can be used by managers to keep things on track.
Security too is an area where this type of analysis is going to gain in importance. The latest behavioural systems are able to analyse the way a person uses a computer or phone and accurately predict when an account sign on is being used by someone else. This has implications not just for securing systems but for policing things like pay walled websites.
There is some scepticism surrounding the value of predictive analytics, much of this surrounds the importance of using data in context. Lots of factors influence individual behaviour and many of these – such as weather conditions or someone having had an argument with their partner - are impossible to factor into even the most sophisticated model.
But despite these doubts business leader are keen to take advantage of the opportunities predictive technology can offer. "The impact of integrating real-time analytics with business operations is immediately apparent to business people because it changes the way they do their jobs," says Jim Sinur, research vice president at Gartner (opens in new tab). "The most dramatic change is the increased visibility in how the company is running and what is happening in its external environment. Individual contributors and managers have more situational awareness, so they are able to make better decisions faster."
Our everyday world is more and more driven by data and although there may be limitations in what predictive systems can achieve, that data can still offer valuable insights. Businesses need to remain competitive and by helping them to identify opportunities, improve customer service and minimise risks predictive analytics is almost certain to play a key role.
Image Credit: Sergey Nivens / Shutterstock