Using Big Data to net online buyers
New Delhi: Imagine you bought a shirt from an online store, but couldn’t find a matching tie. Think about how much easier things would have been if the store offered you tips for suitable ties based on your preferences and what others are buying.
Indian e-commerce companies may soon be able to provide such assistance to customers. They may also be able to suggest upcoming discount offers, in case a customer decides against buying a particular product because he has expended his budget.
Personalized assistance, driven by the intelligence provided by Big Data analytics—churning and processing a large amount of data to draw insights and patterns—is now being touted as the next big thing for e-commerce companies.
Only a handful of companies in India are currently using Big Data analytics to provide customers a basic personalized experience and make recommendations based on the customer’s buying behaviour.
“We haven’t seen too many companies embarking upon this journey, but we are having discussions with some of the big organizations on some use cases (case studies), which are either some newer use cases, specific to the Indian market place or the use cases which have been used elsewhere successfully,” said Asheet Makhija, country leader-information management, International Business Machines Corp. (IBM) India/South Asia.
Apart from personalized recommendations, real-time analytics—responding to a customer’s action in real time—is gaining the attention of online retailers as the primary use case where analytics could be deployed.
“E-commerce companies use Big Data in two ways. One is to analyse past behaviour of customers to find patterns, and the other is real-time analysis, that is, reacting when the customer is shopping online,” said Makhija.
“Similarly, the other thing which some of these companies are looking at is the next best action kind of a step. For example, on one of the travel sites, if you stay on the site and don’t do anything, suddenly an offer pops up saying that Rs.100 off on some offer, because the customer hasn’t acted.”
“Moulding what it is that you want the customer to do as the next best step, these are the things which they are doing (or trying to do) on a real-time basis,” he says. “The deeper you get in each of these use cases, the better it gets.”
Narasimha Jayakumar, chief operating officer, HomeShop18.com, a digital commerce platform of the Network18 Group, said his company tracks the entire customer journey, which includes customer behaviour—how customers look at the site and what products they usually choose, action taken by a customer on the website and how customers navigate the website.
The company has been deploying Big Data analytics since the last 15 months using Hadoop, a software from open source software foundation Apache that can process huge amounts of data.
“We use it (Hadoop) primarily for insights and analytics on how to show right recommendations to our customers,” Jayakumar said.
“A very common use case is recommendation of similar products. For example, if a customer is shopping for a mobile, then one way to give correct recommendation is manually. The other way is, we use a lot of data from millions of customers and visitors to suggest the right sort of phones which are similar. This is what we call crowd sourcing,” he added.
According to Jayakumar, the conversion rate for the company has gone up by 30-40% and the average time a customer spends on the website is up from 2-3 minutes to 5-6 minutes after it began deploying analytics.
Jabong.com is using some “elements” like stacking (combining) data to gain competitive advantage and insight opportunities which have not been discovered before.
“The scope of online retail has reached to the individual and to go to that level of granularity, we need to use Big Data analytics solutions after reaching a certain level,” said Praveen Sinha, co-founder and managing director, Jabong. “We were earlier using SQL (structured query language)-based analytics, but we moved out of it eventually, with changing needs.”
Jabong is currently using analytics for marketing, but eventually, Sinha says, it can be used to drive finance and operational analytics. “We are partly using a system that interacts with Hadoop, but to use Big Data analytics in the entire operations and to take it to the full scale, we would need another year-and-a-half.”
Infosys Ltd agrees with IBM that there has been an active interest from Indian online retailers. “We are getting requests from larger e-commerce companies to help them with their big data strategy at this point. It is similar to what is happening in the rest of the world though the timeline is a little slower when it comes to deployment here,” said Rajeev Nayar, associate vice-president (VP) and head, Big Data practice at Infosys.
“Analytics is only being used by Indian online retailers for front end in the context of personalization, increasing sales and decreasing abandonment,” said Nayar.
“The companies have started using these tools to some extent, but the Indian companies still have not seen happening usage of analytics in understanding competitors’ pricing strategy to get the competitive advantage, which is already happening in the West.”
According to Nayar, Infosys is developing a solution where data from social media and other external data can provide an indication of the problems that may happen in a company’s supply chain.
Snapdeal.com, which has been one of the early movers in Big Data analytics for personalization and recommendation so as to make product searches easier, also said that real-time analytics has become a key focus area for companies.
“The next big thing is real-time analytics, bringing consumers personalized experiences at all times in the same visit. That’s where real juice will come from,” said Ankit Khanna, VP, product management, Snapdeal. “We will be including intelligence such as recommending similar products which have some offer going on or which may come in the next few days. The other thing is, if a customer is shopping on the web with us, we may start sending him notifications on mobile.”
Snapdeal, which has over 20 million subscribers and generates terabytes of data through the interactions that happen with customers in addition to a catalogue of over 5 million, churns 15 million data points (related data set like a consumer shopping on specific days for a particular thing) within two hours, using Hadoop.
About 35% of its orders come from recommendation and personalized systems, and the conversion rate of such orders is 20-30% higher than normal orders, the company claims.
Khanna explains why not many online retail companies are using Big Data: “First, they don’t have that Big Data set. Second, it involves a particular kind of technology platform and you need specialized people to handle that. And third, it costs you significantly so you need to have a good use case to deploy it.”
“The good thing about e-commerce companies, which were established in the last five to six years, is that they don’t have legacy systems and thus can deploy Big Data solutions faster,” he said, adding that Indian (e-commerce) companies are trying to tap the entire cycle from the moment a customer enters the site to the time he or she leaves in order to provide the best experience so as to retain customers.
“Even as the appetite for trying out certain things is better in mature markets, but because of the sheer volume of data, the adoption of big data analytics in some cases is progressing well,” Khanna said, adding that companies “like Flipkart, Myntra and Snapdeal have already achieved operational efficiencies and are getting into it (analytics)”.
IBM’s Makhija said over the last two to three years, e-commerce companies in India have seen phenomenal growth, but are grappling with the basic issues right now. He added that many companies will deploy Big Data analytics in the next 12-24 months “because by that time, they would have felt the market place and competition as well as realized the importance of getting deeper into knowing a customer, having a targeted campaign and, increasing the value per customer.”
This is the fifth in a seven-part series.