Flipkart harnesses 10-year data to build AI platform for sales prediction

E-commerce major Has an advantage of its 100 mn user base, majority of them using its smartphone app

As India's largest e-commerce marketplace Flipkart closes in on completing a decade in the business, it is looking to put in use its mammoth pile of data to predict sales of products months in advance.

The company is working on an artificial intelligence (AI) solution that will give it an edge over rivals by helping it make smarter decisions in ordering, distribution and pricing products on its platform. Ultimately, the AI system will allow Flipkart to boost efficiency and reduce the cost of products for customers.

"We (Flipkart) are trying to predict how many units of what we will sell. This is a very complex thing that is dependent on a variety of inputs such as price, discount or if an event, such as Diwali or Christmas is coming up," said Krishnendu Chaudhury, principal scientist and head of image sciences at Flipkart.

While rival Amazon, which has around a 10-year headstart over Flipkart, is known to have some of the most advanced sales prediction engines, the Indian company has the advantage of having a bigger data set of the country's online consumer market.

For AI engines to be trained, the more data available, the more accurate researchers can make them, so Amazon's advances globally might not count for much in India as consumer behaviour varies depending on the region as well as the shopping season.

Flipkart, among the early entrants into India's e-commerce space, also has an advantage of its 100 million user base, the majority of them using its smartphone app. This gives it better insights to understand consumer behaviour, including their socio-economic background based on the smartphone they use, the signals they can capture from the places they travel and can target with the right products.

It also helps that for majority of Indians, a smartphone is their first computing device, through which they listen to music, stream videos, share photos and transact online.

Flipkart has not yet applied for a patent on its AI sales prediction engine and hence isn't keen on sharing too many details to safeguard its technology. The company says it has been successfully using AI in the fields of computer vision and language processing to build tools that help customers better search for products on its platform.

"It (Flipkart's AI engine) will also look at something like how much hype is being created in the media about a certain product. For example, iPhone sales depend hugely on the media hype. So we have to capture all these various things and it will make a very complex decision," added Chaudhury.