Customer Analytics at Flipkart.Com It was typical cloudy monsoon weather at Bang

Question

Customer Analytics at Flipkart.Com

It was typical cloudy monsoon weather at Bangalore on July 28, 2015. In the Darwin room of Flipkart’s Cessna Business Park office, Ravi Vijayaraghavan, the head of analytics and Pravin Shinde, senior manager Analytics were brainstorming various business problems that Flipkart as an e-commerce company was encountering. Flipkart had been putting in much effort and emphasis on the use of Analytics in every aspect of decision making. Forecasting demand for thousands of stock-keeping units (SKUs), predicting returns and cancellations of orders, predicting the reasons when customers contact the customer service centers, optimizing markdown pricing, identifying various types of frauds, optimizing vehicle routing, and enabling adherence to service-level agreements, were some of the typical problems that the analytics division of Flipkart was solving using state-of-the-art analytics techniques. In 2015, the team included about 100 data scientists mostly recruited from institutes such as the Indian Institute of Technology and Indian Institute of Management specifically for this purpose.

E-commerce in India had seen a compound annual growth rate (CAGR) of 34% since 2009 and was expected to exceed USD 22 billion by 2015. Under the e-commerce head, e-travel in itself comprised 71% of the total e-commerce market, e-tailing which comprised online retail, and online marketplaces have been growing exponentially and well-poised to become the fastest growing segment, expected to reach USD 56 billion by 2023. The industry believed that growth was at an inflection point with the key drivers being broadband internet, rising standards of living, wider product range, and changing lifestyles of Indian consumers. Such high growth rate also created several business challenges to e-commerce companies as well as to their market place suppliers, among them profitability still remained a major challenge. The e-commerce companies in India incurred combined losses of around INR 10 billion through heavy discounting to penetrate into the brick and mortar retail customer base.

Ravi Vijayaraghavan started the meeting by stating:

We have been analyzing our data to gain insights, but, do we know the value of our customers? I think it is important for us to differentiate our customers through metrics such as customer lifetime value, which will help us to manage them effectively. For example, we can make our promotions effective if we know the customers with high customer lifetime value.

Customer lifetime value (CLV) is the net present value (NPV) of future cash flows (or profit). CLV is usually calculated at a customer segment level. The main challenge in calculating the lifetime value of customers of e-commerce companies such as Flipkart is that the exact life of the customer is unknown owing to data truncation; that is, the actual point in time of customer churn, may not be identified in ecommerce, since there would be no prior communication from the customer about the churn. Hence, traditional models of CLV calculation may not be appropriate for e-commerce companies such as Flipkart.

ABOUT FLIPKART

Flipkart, the poster child of Indian e-commerce, was an early entrant in the nascent Indian e-commerce market and quickly established itself as the leading company in this space. It was founded in 2007 by Sachin Bansal and Binny Bansal, both alumni of the Indian Institute Technology, Delhi. They pooled in INR 2,00,000 (approximately USD 3,150) each to start Flipkart in 2007. From a startup with an investment of just INR 4,00,00 (approximately USD 6,300), Flipkart had grown into an online retail giant, valued at over USD 15.2 billion as of 2015. Flipkart was running the marathon with ample support from private equity players such as Tiger Global, which invested over USD 1 billion as of 2015. 4 Flipkart sold over 30 million products from more than 50,000 sellers in 70+ categories and consisted of 30 exclusive brand associations, with in-a-day guarantee in 50 cities and same-day guarantee in 13 cities. Flipkart was 33,000 people strong and had over 50 million registered users with over 10 million daily visits and 8 million shipments per month. A burgeoning consumer class, coupled with a rising web literate population and zealous venture capital funding propelled Flipkart to become India's answer to Alibaba and Amazon.

The use of e-commerce to buy products and services was spreading at a fairly rapid pace in the psyche of the Indian consumer. In Indian cities such as Bangalore, lack of time, ease of shopping, and attractive pricing were major drivers for online shopping. On the other hand, accessibility to a variety of products encouraged customers from smaller towns and cities to opt for the online route.

CUSTOMER ANALYTICS AT FLIPKART

E-commerce companies such as Flipkart had access to huge amount of data, available for applying predictive, and prescriptive analytics to take data-driven decisions. Flipkart had a strong analytics team headed by Ravi Vijayaraghavan, which used statistical models and machine learning algorithms to generate crucial customer insights. Flipkart had more than 50 million registered users and the transaction data of these customers could be used in a far more meaningful way using analytics to predict online consumer behavior.

In 2015, the Indian e-commerce market space was facing immense competition owing to the entry of Jabong, HomeStop18, Infibeam, Indiaplaza, Snapdeal, and a plethora of other pure-play and multichannel e-commerce companies. The continued growth of e-commerce and tough competition compelled Flipkart to seek a competitive advantage through more sophisticated analytics. Flipkart has been taking several analytics-enabled decisions, for instance, using web analytics to determine which landing pages encourage customers to make a purchase as well as which pay per click ad campaigns were most effective. In the face of tremendous competition, more than ever before, the analytics team at Flipkart aimed to predict customer demand for the products, understand its customer’s loyalty, assess the true impact of customer retention strategies (discounts, coupons, and extra services), and focus on customer segments with higher retention and spend potential.

In 2015, Flipkart wanted to understand its customers better and retain most of them through effective promotions, since customer retention is less expensive as compared to customer acquisition. Unlike the churn in the telecom sector, which was clearly defined and captured (in the instance of postpaid customers), churn for e-commerce companies was difficult to define and capture, as these events were unobserved. Across e-commerce companies, the customer churn may be very high owing to reasons such as need fulfilment, cessation of demand, competition, and so on. However, it was important to capture customer churn and identify which customers should be retained.

CHURN ANALYSIS AND LIFETIME VALUE

E-commerce companies faced a scenario of inconsistent customer purchase pattern wherein the gap between purchases could stretch far more than 6 months. Even though most of these buyers could come back after a gap of 5–6 months, Flipkart aimed to identify high value customers, and subsequently increase purchase traction among them.

The analytics team at Flipkart wanted to model customer purchase patterns, repeat buyer trends and calculate churn probabilities to help them identify the repeat customer segment to focus more on these customers for their marketing and promotional strategies. The final objectives of this exercise were to forecast the revenue generated from existing customers and calculate their lifetime value.

DATA DESCRIPTION

In order to carry out a detailed customer value assessment addressing customer churn issues, Flipkart collected sample transactional data spanning across 2 years: January 2013 to December 2014 such that all the 30,000 customers in the sample had made at least one purchase in January 2013. This was done to ensure new customers in the above said period were excluded from the study.

DATA ANALYSIS

To understand customer churn and lifetime value, Pravin’s team decided to use Discrete Time Markov Chains (DTMC). To build the churn model, the team first had to identify the period of inactivity (gap between transactions) to define churn. Gap was thus defined as the difference in months between two successive purchases or the difference between the current month (despite no purchase) and the last purchase month.

Pravin’s team also wanted to forecast revenue from existing customers and therefore built a model using Recency (defined as when was the last purchase was made by the customer) and Monetary (defined as how much money was spent in the latest month which had a purchase). The team retained the recency state definitions and augmented the state space by adding monetary slabs for each recency level.Whenever a customer is in an inactive state, the monetary values are retained from the month of the last purchase. The TPM for the Recency–Monetary DTMC is provided as an accompanying file to the case.

The team was also keen to identify customer segments based on their current transactional attributes such as recency, monetary, and frequency. The objective was to provide a fairly accurate mechanism to study the purchase patterns of various customer segments and thereby enable effective promotions to increase customer spend and arrest customer churn. The team chose a quarterly transition time period to account for macro-level stochastic changes in the model. Whenever a customer is in an inactive state, the frequency and monetary values are retained from the quarter of the last purchase.

The transition probability matrices from each of the models were used to calculate the customer lifetime value for customers in each segment and subsequently build an effective campaign strategy to reduce churn and increase customer spend.

Outside sources may be needed. Define the place characteristics of the marketing mix.

Place:

Objectives:

Type of channel (Direct, Indirect, Mixed):

Wholesalers/Retailers:

How discrepancies will be handled:

How are functions shared:

Coordination needed (information system also):

Transportation requirements:

Inventory product-handling requirements:

Facilities required (private/public warehousing, distribution centers):

Reverse Channels (Returns, recalls):

Explanation / Answer

Objective: “Place” in marketing mix of a company aims to deliver products to the end user in right time and place.

Types of Channels:

Direct Channels: The manufacturer directly deliver the product to the customer in this channel. In this channel the company have control over the product and user’s feedback.

Indirect channels: This channel is needed by the manufactures for huge sell of the product through small retailers and wholesalers, it increases the cost of the product, as because the middle man between the costumer and manufacturer will keep his profit.

Mixed Channels: This channel is basically consisting of Direct + Indirect channels. In this channel the company can sell their product in either ways: May be directly from manufacturer to the customer or Manufacturer to the customer through Retailers or whole sellers.

Wholesaler/Retailer: If a manufacturer wants to sell its product widely, then he can sell it through middle man/third party who are also known as Wholesaler/Retailer. The Retailer/Wholesaler have huge number of contacts by which they can sell the product by keeping their profit to the customers. When these Retailers or wholesalers get the product, they may request to omit the name of the manufacturer company and it is the Retailers strategy to build relationship with the customers.

How discrepancies will be handled: The discrepancies can be handled by:
a) Identifying and verifying the order placed
b) Checking for the same product at different locations
c) Checking for the merge of similar wrapped product
d) Checking the data processing errors
e) Counting the product

Reverse Channels: In this channel the customer returns the product to the Manufacturer/Wholesaler and get back his money if he is not happy with the product delivered to him.

Related Questions

Navigate

• Browse (All)
• Browse C
• Subjects

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.