By 2020, the world would have generated around 40 zettabytes of data, or 5,127 gigabytes per individual, according to an estimate by research firm International Data Corp.
). With data being touted as the next oil, companies have begun sharpening their focus on analysing this deluge of data to understand consumer behaviour patterns that could help them drive growth further, agreed industry veterans and experts during a panel discussion on What should CIOs do to make business sense of the deluge of data
The panel comprised Satish Mittal
, chief technology officer, Vodafone Business Services
; Sandeep Dhar
, chief executive officer of Tesco Hindustan Service Centre
; Sanket Atal
, chief technology officer of MakeMyTrip.com
; Amitabh Misra
, vice-president, engineering, at Snapdeal.com
; Srikanth Velamakanni
, chief executive officer of Fractal Analytics Inc.
and Amit Khanna
, partner-analytics, KPMG India
. The discussion was moderated by Mint
’s technology editor Leslie D’Monte. Edited excerpts:
What are telecom services providers doing with this huge data influx?
Data is the new currency. The question is, is it happening? Is it real? For Vodafone
, it is. For instance, we have joined hands in the Netherlands and Turkey with town planners, whereby we give them the entire analytics of how traffic is moving since every moving vehicle has a person who has a mobile phone. So we can easily identify the pockets where traffic is moving, where people are going, and intelligently divert traffic to avoid congestion points. Another potential use that is being worked out is with ATMs. If we can combine information of an ATM swipe of a particular location with the registered mobile of the user, you will know whether that person is at that location and whether the person who is using the ATM and the person who is registered as mobile user are the same. This will reduce the chances of a fraud. As an enterprise, we can use data to analyze how customer experience is during certain times of the day in certain pockets—how many of them are, for instance, iPhone
users or Android
users—and give this feedback so that manufacturers can address these problems when designing the phones.
The retail industry and e-commerce firms also love data. It’s crucial to their survival given the stiff competition.
Dhar: Every time a customer buys at one of our shops, he wins loyalty points and because there is an incentive, he identifies himself. With this mechanism, we are able to link all sales to individual customers. This is what happens with analytics, whether it is an enterprise data or data in social space—you go through the cycle of predicting, personalizing and planning.
For example, if a shop found that a lady has suddenly started buying products that are suitable for expecting mothers, it is an opportunity to enrol her into expecting mothers’ club and send targeted offers. We can even send her literature, which an expecting mother would find useful. Starting with this, you can actually have a comprehensive plan around life-stage marketing. You will now be able to predict when the baby comes into the world, when the baby is going to go to school so that you can offer range of products for first time student—you can actually track the entire life of the baby and at appropriate times offer the family home loans, student loans and so on and so forth.
We have analytics running in our veins. Snapdeal
is a technology platform that brings together 20 thousand sellers and over 20 million subscribers. But only a handful of employees, in about thousands, manage this interaction. If the data was in raw form, this would have been impossible since data generated per day runs into terabytes. So we have a very sophisticated layer of intelligence on the top of data. Every single decision that we make internally is based on that. About 31% of our orders come through our analytics-driven systems. We record the buying behaviour of buyers and customize things for them.
If you visit our homepage, we have personalized and recommendation pages. Similarly, if you look at our search, you may think that when you are searching a key word, the set of products that will show up is going to be constant. That’s not the case. Based on millions of people searching a product, the result of relevance and the form in which search is shown to users, keep improving. Based on the selling and buying patterns, the segmentation and profile of buyers, all these numbers are crunched to personalize an email.
Travel is another industry where there is a huge amount of unstructured data.
Atal: It is very important to understand the mind of the customers. Big data analytics—a combination of structured and unstructured data—gives us the insight into what’s going on, which is sometimes not obvious.
For instance, there was a retail entity trying to figure out what would be good to put next to each other to increase sales and one thing they came up was diapers and beer. When they did that, sales of both the items actually went up. Our analytical platform is also on the top of the vast amounts of data that we have in-house, but that’s not a social data. Our requirement is traditional data as well as realtime data.
We try to understand everything about customers—from basic things like their travel patterns, the kind of hotels they like to stay in, who are their typical co-travellers, their experiences, etc. All this is geared towards getting us a persona of a customer identified internally, so that when the customer comes to our site, he can have a very personalized experience. If a customer searches for hotels in Goa, we should know whether he likes 5-star hotels or not, what kind of events the customer is interested in, what kind of experience customer had in relations to hotels. And then the sequence of hotels that we show, it should be in the sequence which is conducive to the customer choosing from first five choices.
From excel sheets to Big Data, how have the tools used for analytics changed?
Velamakanni: The tools have changed dramatically because the size and complexity of data has changed.
When we started out, we had basically model-building platforms (SAS, etc.) that were not necessarily very open. Tools today have become much more useful because you are able to integrate them. On the infrastructure side, we have big data related-technologies like Hadoop
. The other thing we have seen becoming really useful is rapid data discovery, for which we have tools like Qlikview, Tableau, Spotfire and Domo.
Now that the data complexity has multiplied dramatically, we need to do visual stories and simplify things to see that what’s going on is still important. In general, tools come and go, but what is more critical to understand is—what is the business problem you are really solving? If you are trying machine learning-driven personalization of a page, for instance, you can’t just build a SAS-based predictive model and then try to deploy it. The system has to learn with every new transaction. None of the tools I described might be able to do that job. You might have to write some code after you figure out how to learn from this data.
Is this experience across sectors?
Khanna: Big data analytics is definitely a buzzword. It is the next wave of how businesses are going to be conducted because people are getting more factual and they need everything to be driven by numbers. Everybody is talking about analytics.
The underlined word here is talking about analytics. We recently talked to 170 chief executives globally. Sixty eight percent of them didn’t know what they were doing with the data.
In India, I went to 50 top companies and went to top five heads in these companies. We asked all of them on the second of the month—What was your last month’s sales? For 72% of the companies, the five answers were different. If we call this analytics, it is not. The point is, analytics is not about technology or gather a lot of data and doing all this high-funda stuff. It is about how you can use the data which you have collected more effectively for your business. If you can’t do this, all the tools and what you are trying to do with them is not useful.
The story has been modified from its original version to clarify the name of some analytics tools.