Tuesday, March 4, 2014

Flipkart's 'Demand Sensing' Innovation

A few years ago, shopping online was a big deal in India. One of the major reasons behind this was the need for a Credit Card to do so. I remember the first time I tried shopping online in India and being disappointed on being asked for a CC. I had the money to pay for the goods, just not the plastic form of it. This was a problem faced by many Indians. Even those who owned a credit card, were afraid to use it due to the CC related frauds which were all very well publicised in the news.
All this changed when the e-commerce site, flipkart started its operations in 2007. They had a plan. By accepting “Cash on Delivery” mode of payment, they planned on targeting the very problem that stopped people from shopping online. When rapid advances in technology were encouraging everyone to use plastic money, flipkart took a step behind and started dealing in cash. And it worked for them. It was a huge risk in terms of dispatching products without having actually received payment, but it paid off. Today flipkart is the largest e-commerce site in India grossing revenues of more than 190 million as of Dec, 2013.

This journey of flipkart has not been easy, to say the least. Starting an e commerce site in India has posed its own set of problems which Flipkart had to tackle. The robust back-end that they built played a vital role in their success. Like most other e-commerce businesses flipkart started off with the consignment model of business, in which they acquired goods from suppliers on scoring customer orders. As their business grew, they shifted to the warehouse model, managing their own inventory based on consumer demand. With the need to manage an inventory came the need to forecast customer demand. Since they had no historical data to rely upon, they came up with their own innovative solution to use real time demand sensing and website performance measuring framework.

The VP of engineering, Amod Malviya came up with a two part real time website performance measurement framework. The first part of this framework has certain metrics which measures the performance of the website in the real time. The second part includes the A/B method of testing the website. The A/B method of testing is a type of hypothesis testing. By having 2 versions of the same website, differing in a certain aspect ‘x', it lets analysts infer what the impact of the change ’x' has on the website. This has allowed engineers/analysts to instantly deploy new models and test them instead of wasting time debating on whether they'll be effective.

Flipkart has the capability of having multiple live versions of the website at the same time. Whenever, an engineer/analyst wants to test a new model, he can divert a certain percentage of the web traffic, say 10%, to the new website model and measure its impact on sales. The website performance metrics measure the changes in performance in real time and throw alerts if the sales fall below a certain threshold. These changes can then be rolled back. 

This strategy has also been a huge hit with suppliers who can float new products in the market and get a sense of the demand before introducing them in earnest - a new and innovative way of sensing market demand.

Flipkart was started with a modest invest of $8000 in 2007 and since then has become a multi million dollar enterprise. In a way, company’s like Amazon and Ebay have shown Flipkart the way forward. But now that they are equals in the field of e-commerce in India, flipkart faces tough competition from these rivals. For flipkart, constant innovation is the only way to survive in the world of retail e-commerce.

Flipkart drives innovation through intelligent use of IT.
A/B testing:

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.