AI Tackles the Indian Parcel Problem

Flipkart Uses Artificial Intelligence to Set the Record Straight

India’s e-commerce market is still settling in. Just over the past few days, the takeover battle for the top company in this space, Flipkart, has been intensifying.

Walmart is already in advanced talks to acquire a 51 percent stake in the Indian retailer, but Amazon understands the stakes and is expected to make its own offer.

While this merger and acquisition activity is happening in the background, Flipkart is working to improve its services by tackling problems with the tricky Indian postal address system, and by detecting and preventing the growing threat of address fraud. Flipkart’s data scientists have developed an innovative solution that relies on artificial intelligence to correctly identify addresses in the country. According to the company, the system resolves address inconsistencies with a 98 percent accuracy rate.

To understand why something that might seem as straightforward as finding a postal address would need an intricate artificial intelligence system, one has to understand India’s unique postal address logic. It really can take the power of a supercomputer simply to figure out where to send a package.

In the absence of a formal geospatial classification based on latitude and longitude coordinates, postal addresses in India are organised by Postal Index Number, or PIN. However, each individual PIN code may correspond to an area as large as 50 square kilometres. The PIN system, which was implemented in 1972, is inadequate to serve modern online shopping consumers.

This outdated and complicated system is causing a huge challenge for last-mile logistics. Incorrect addresses not only lead to delayed or failed delivery of shipments, but also severely affect the revenue models of online retailers and have an adverse impact on customer satisfaction.

Flipkart, which deals with millions of strange addresses, has found a vast number of spelling variations after analysing the addresses using its system. In an ideal world, postal addresses are required to be consistent to enable correct and efficient delivery of mail. They should be structured in a hierarchical manner, for example, by country, state, district, city or town, locality and so on. However, in India, they’re neither consistent nor do they accurately convey or represent geolocation.

Flipkart is also hoping to solve another distinct challenge in the country: consumers and businesses looking to take advantage of extreme price discounts of limited-supply offerings. Resellers exploit the online discounts that Flipkart offers and sell products offline for profit. They’re very difficult to pinpoint, but the company has built machine learning models to identify them.

Flipkart has more than 100 million registered customers and has learnt a great deal about Indian shoppers. As its reach expands to new demographics, the company has developed a complex system to address a complex problem. It has hired top talent in artificial intelligence to develop ways of identifying the correct delivery locations as well as deterring fraudulent transactions. Artificial intelligence is finding an important place in the Indian e-commerce arena.

Sign up to our free Daily Insight service here.