The power of predictive analytics

Andy Bauer, Business Connexion, says predictive analytics can even take into account the impact of floods in China on a particular organisation's supply chain.

---

A recent report by Gartner states that organisations that use predictive business performance metrics will increase their profitability by 20 percent by 2017. The research house maintains organisations should use predictive metrics to alert workers that a business moment is about to occur, and guide them on the next action to take in the context of a particular customer's expectations.

"Using historical measures to gauge business and process performance is a thing of the past," says Samantha Searle, research analyst at Gartner. "To prevail in challenging market conditions, businesses need predictive metrics - also known as 'leading indicators' - rather than just historical metrics (aka 'lagging indicators')."

She then makes the bold statement that: "Business process directors who don't apply predictive metrics to cross-boundary business processes will leave their organisations vulnerable to the risk of failing to execute their business strategies."

These are certainly unequivocal statements in support of predictive analytics. So Brainstorm spoke to three senior IT decision-makers about their views on whether these metrics are indeed vital to the future competitiveness of companies.

Andy Brauer, the CTO of Business Connexion, agrees that predictive analytics is critical to business success. "Where companies get it wrong is that they look at business intelligence - historical information to see what happened. They're not taking the next step, which is to look at why."

He raises the point that 95 percent of business cases fail, indicating that while we're flooded with information, we're starved for knowledge.

"Predictive analytics makes use of the Montecarlo Principle (a mathematical technique that was originally used to predict gambling table performance) to work out what's most likely to happen - to see patterns emerging before they become trends, and to experiment with multiple scenarios before making a decision."

In this way, he says, predictive analytics really gives organisations the proper metrics for where they're going and what they want to achieve, blending data from inside and outside the organisation. "It can even take into account the impact of floods in China on a particular organisation's supply chain."

Richard Samson, KPMG's Africa CIO, is confident of the strategic and competitive benefits that are offered by predictive analytics. He referred us to Frank Rizzo, the partner who leads the data and analytics team at KPMG in Africa.

Rizzo says KPMG in South Africa is assisting a number of clients with predictive and descriptive analytics solutions. "The nature of the solution is dependent on the particular client's maturity with analytics and the quality of their data," he says. "My view is that this requirement will increase over time as clients start analysing more and more structured and unstructured data."

Historic measures can still be relied on where predictive metrics can't be used, aren't appropriate or feasible.

Lourens Walters, Medscheme

He agrees with Gartner's view that technologies depending solely on historical measures are a thing of the past, and that predictive analytics solutions will ensure a competitive edge. This, he says, is in line with KPMG's global research that was published in a paper titled 'Going beyond the data: Achieving actionable insights with data and analytics'.

One of the conclusions in the paper is that: "Those organisations [that are] able to connect the dots will be poised to achieve quantum benefits from their D&A [data and analytics]. Those that cannot - or those that do nothing - will be forever submerged in data, with its value untapped."

Lourens Walters, GM of the Health Intelligence Unit at Medscheme, also agrees with the Gartner predictions, with certain provisos. "I endorse their statements, but being a scientist, I must stipulate that I endorse them only if substantiated. Where appropriate, predictive metrics will supply a competitive edge, but they're not appropriate in all cases. Historic measures can still be relied on where predictive metrics can't be used, or aren't appropriate or feasible."

He adds that he has absolute faith in predictive analytics, because it's possible to test if it's accurate or not by running the solution on historical data.

The Health Intelligence Unit at Medscheme has been using predictive modelling or advanced analytics for the past 19 years. "These are used to analyse healthcare-related data. The data are collated from various data servers used in the Medscheme environment into a dimensional data mart on the system used by the unit, which is then exploited by business and analyst users alike," says Walters.

One aspect of the system's predictive work identified individuals within lower morbidity groups - essentially, who had less life-threatening conditions - who were most likely to incur increasing healthcare costs in future, but would still be amenable to care management intervention programmes. The result of this was financial savings and a reduction in utilisation of hospital services. For this research, Medscheme won the Starfield Award from the Johns Hopkins University in the US.

While the general enthusiasm for the benefits offered by predictive analytics among this survey's respondents was high, it must be noted that the decision-makers who were willing to discuss the topic were those who are involved in some kind of implementation. Given the reticence of those whose organisations are not currently using predictive analytics, the market would be wise to heed Walters' words: "Where appropriate, predictive metrics will supply a competitive edge, but they are not appropriate in all cases."

First published in the April 2014 issue of ITWeb Brainstorm magazine.