After the user consumes the tea, the company requests for feedback based on which the recommendations are fine-tuned. Using previous behaviour and different data points captured from a user to predict what he or she is more likely to buy is not new in ecommerce. For the tea industry, where sprawling estates still run on ledgers designed in the British colonial era, this is right out of the future. "Tastes are unique and with each prediction our engine gets better," said Dugar. "With the first order, we are 75% sure of which teas you are going to like. The next order has85% accuracy. As you taste, read and go through sessions, you become a connoisseur," he said. Teabox, founded in 2012, uses technology to improve the tea industry's supply chain. "The quality of tea deteriorates due to lack of storage and transport infrastructure. We take care of that by being at the source, shipping tea within 48 hours of production and reducing the time for consumption to 5-7 days," said Dugar, who hails from a family of traders who supply equipment to tea estates. One of the company's early backers, Oak Hill Capital founder Robert Bass, was originally a customer who was impressed by the freshness of the tea sold by Teabox. .
Siliguri-based Teabox is launching an algorithmic prediction engine to help novice tea drinkers figure out whether to go for the Darjeeling black first flush from Jungpana plantations or something else based on their responses to a few basic questions. "It's very much like wine (where) (where) when you start you probably know only red and white. There are so many varieties of tea (making it tough for one to choose). We came up with the prediction engine to solve this problem," Kaushal Dugar, founder and CEO of Teabox, told ET. The venture backed startup's new service works like how Facebook would personalise your news feed based on what its algorithms think you like most. Each new user on Teabox is asked five basic questions such as what kind of chocolates and smells they like, based on which the recommendation engine assigns a unique signature to the user. Then it recommends a tea after taking into account nearly 75 different attributes including aroma, astringency and strength to match the user's liking.