The Hindu Business Line : How ‘predictive’ can you get?

Often predictive analytics is confused with business intelligence. Are they the same or different? If different, how so? Tune in to the Head of Predictive Analytics, Persistent Systems, Dr Kedar Swadi, who holds a doctorate in computer science from Pr inceton University and is researcher at Rice University. Excerpts from a chat with eWorld.

How are enterprises taking advantage of predictive analytics?

This concept is playing an increasingly important role in enterprise decision management systems. While traditional Business Intelligence provides a retrospective and aggregated understanding of data, Predictive Analytics allows enterprises to extract latent information present in their data and derive a prescriptive and actionable understanding that can be applied at the level of each and every business transaction.

Businesses, across domains, have based their decision management systems on Predictive Analytics-based solutions to arrive at informed decisions faster.

For example, in the retail domain, this concept is used for item-specific inventory control systems. In marketing, it allows enterprises to extract the most promising target groups from individual customer data and maximise marketing campaign revenues. Financial and investment firms have deployed high-accuracy Predictive Analytics-based solutions to identify potential risks and frauds in time to prevent them. Life Sciences is another area that finds applications for this concept, from drug discovery, to healthcare management, to analysis of gene expressions.

On trends in this space…

We see an accelerated drive for adoption of Predictive Analytics solutions. According to a recent survey by TDWI Research, about 40 per cent of large enterprises have some form of analytics solutions either implemented or under implementation. Those not in this group are actively exploring its adoption in their decision management systems. While a large number of current applications are related to customer relationship management, financial planning and risk prediction, we see a rapidly growing interest in Web analytics, whereby businesses are able to tailor online content based on the profile of persons visiting their Web site, thus extending the value of their online establishments.

How is Predictive Analytics built on Business Intelligence solutions?

It would be incorrect to say that Predictive Analytics is built on Business Intelligence solutions. We see these two as complementary techniques, each of which has its specific role in enterprise decision management systems.

Business Intelligence has traditionally been used for aggregation-based retrospective analysis, which would help analyse data across time periods, geographical areas, administrative units, etc. Predictive Analytics is suited to discover latent patterns and exploit this information to be able to generate high-accuracy predictions, and prescribe actions based on these predictions.

In the past few years, Business Intelligence solution providers have started incorporating Predictive Analytics features in their software. Not only does this make their systems more integrated, it also makes Predictive Analytics techniques more accessible to engineers who are used to traditional Business Intelligence tools. We soon expect to have most Business Intelligence vendors providing a seamless integration of Predictive Analytics techniques in their tools.

What is the current market size? In which sectors is Predictive Analytics witnessing traction?

According to IDC, the worldwide revenues from Predictive Analytics-based activities will top $3 billion, with a compounded annual growth rate of around 8 per cent.

This concept is currently witnessing most traction across multiple domains, from demand in CRM (customer relationship management), sales, marketing and Web analytic applications. There is also traction in the BFSI sector, especially for forecasting, planning and risk analytic applications. Pharma and law enforcement are also witnessing an increased interest and spending on these systems. A few years ago, there were a handful of niche vendors supplying Predictive Analytics solutions. Today, we see not just these companies, but medium and small-sized enterprises entering this field. This can be credited not only to the maturity of both the theoretical foundations and practical implementations of techniques but also a growing awareness of Predictive Analytics as a discipline.

Another differentiator is the fact that personnel required to set up Predictive Analytics practices need a different skill set from that of a typical software engineer. They need strong mathematical and statistical background, experience in its tool techniques as well as some familiarity with traditional software engineering. In addition, it is necessary to develop a deep understanding in one or more of the important domains for which we develop solutions.

vrishi@thehindu.co.in