Thought Leaders in Big Data: Interview with K.R. Sanjiv, SVP Analytics and Information Management Services, Wipro (Part 2)

Sramana Mitra: It is definitely not a volume game. This kind of volume is well within the capacity of data analysis. This has been around for a while. So why do you categorize it as a big data problem, and what is it that you are specifically doing in those circumstances?

K.R. Sanjiv: I’d like to step back to answer your question – in terms of what big data is. Sometimes it gets intriguing what is big data and what is small data. Strictly speaking, the concept of big data is not a new asset, one that the industry has found today. The data in its form already existed, but the data was always larger than the information systems could handle to derive useful data out of them. What has happened over the past several years is that it has exploded. The digitized information has exploded from a volume perspective and also from a business perspective. Big data is not a new aisle. It is just that the existing data has exploded at an exponential rate. There are new data sources that are coming in, and that is probably the new application area. But from a typical enterprise perspective, it is about volumes being handled in real time and faster.

We did an economic survey. What that survey said is very interesting. The difference between high-growth and low-growth companies is not that some collect data and others don’t. Almost 55% of the low-growth firms collect the data, but they do not use it downstream to get insights as well as the high-growth companies do.  So, the difference is in how you are using the data and how fast you can build the data.

SM: Let me offer a slightly different point of view on how I look at big data versus regular data. The volume of data that is available today is much higher and is being generated on a continuous basis through mobile and social, people doing stuff constantly, enterprise networks, etc. There is an explosion of digital activity all over the place. That is one vector. The other vector is that people who are actually deriving meaningful insights out of it and doing the most interesting things with that data are the people who apply machine learning to develop heuristics and actionable insights – not just analyzing but developing learning algorithms to model that data and derive actionable insights out of it. That is where I am looking for innovative projects you are engaged in.

KS: What you are saying is absolutely right. It is not necessarily just a big data problem. Analytics will change drastically over the next couple of years in two ways. One is about more self-learning, more intelligent systems, etc. We already see that happening. It is not just happening in big data, but also in regular data. For example, if a retailer runs a campaign, the results of that campaign will be automatically fed back into the system, which learns. I can give you a good example of where the data is being used. A self-learning system has been implemented in analytics. This is for an automobile company selling used cars. The dealer has half a minute at an auction to decide whether to take the price or not. If he takes the price, he leaves money on the table if it is lower quoted, and if he doesn’t accept the price he is probably getting unnecessary inventory cost. There is an analytical system built to give a recommended price. If the actual deal is more or less the recommended price, it feeds it back with a reason. It then takes this reason into account in subsequent recommendations.

The same happens in an investment bank for the advisors. You give advice and if it is not accepted, the reason is fed back for learning. So, self-learning in analytics is definitely the next generation thing. Then there is a second change to analytics – and this is going to happen much more industrialized. Today most analytical applications run as departmental silos. Where the shift will happen is the move from data to information. This is a result of processes getting re-engineered and technology stacks getting redesigned. The focus will shift from data warehouses that people built to information warehouses for analytics. If you are a telco provider, for example, information looks like “there have been three call drops in the last minute” – that is information – or “somebody has withdrawn twice the average amount from an ATM than usual.” Analytics requires information. The shift will happen into these information warehouses rather than data warehouses.

These are two shifts we have seen in the analytics space, and this is triggered by the amount of possibilities in big data. We see projects in the emerging area of manufacturing and intelligent machines. We see a lot of analytical systems being used for the M2M – machine to man – environment. There are man-to-man, man-to-machine and now also machine-to-man applications coming in. In the future there will also be machine-to-machine applications, but that is still more or less five years away. Today you have machines recommending to man what shall be done to manage machines more effectively – how the yield of a batch can be increased in a manufacturing environment, for example.

This segment is part 2 in the series : Thought Leaders in Big Data: Interview with K.R. Sanjiv, SVP Analytics and Information Management Services, Wipro
1 2 3 4 5 6