Here’s how Meesho uses data and AI for price optimization - ET CIO (2024)

• Next-Gen Technologies
• 3 min read

The right price could make or break a business. Underprice the product and you offer too much on the table, overprice it and you can't sell. You certainly can't enter a price war with your competitors but also can't make a guess. Businesses don't work anymore on guesses, they work on data.

e-commerce platform Meesho was faced with a similar challenge and tried to leverage data and modern technology as a solution.

“With the Jio revolution in India, a lot of people came online. Their requirements and needs are very different from tier 1 or 2 city audiences, extremely value conscious and not brand-oriented but they are interested in the latest fashion coming up. Off lately the trend in e-commerce space has changed from intent-based to browsing-led. So people just keep looking at things hoping they would like something,” explained Sanjeev Barnwal, Co-founder and CTO, Meesho.

Unlike Flipkart and Amazon, Meesho operates in the unbranded market where the consumer would rather save 10 rupees on a purchase than pay for the next-day delivery. Dealing with a number of manufacturers, retailers, and designers in the unbranded space states that they don't have benchmarks on how to price the product. And it was important for the users to get just the right price.

“If the supplier lists the product at a wrong price, then it will not sell enough which is ultimately not good for Messho as a platform. For sure the supplier will not be successful but also the demand will be left unfulfilled,” he said.

Meesho always had the data-first mindset. This mindset helped the business solve this problem of price matching.

The company invested in a price recommendation algorithm, an AI lead system that looks at the history of Meesho and product sales to suggest a price to the supplier. It tries to understand the product, maps it to certain similar products, and then understands what price works best for the product.

The algorithm looks into the past data as what similar products were sold at what price and what products were not sold at what price. Let's say if a similar product was priced at X and sold Y units but when it was priced X-50, it sold 10Y units. These are the relationships this algorithm makes to come up with a price for the asked product.

The price recommendation model has a good response from the supplier side. Close to 90% of the suppliers have tried and are using this to list their products at the best price.

Sorting the data problem

Because the businesses operating with Meesho are small with limited expertise and human capital, they don't know how to provide the data in a structured manner. And it was wrong for the company to ask these suppliers to upload a bunch of excel sheets. Because of this, the coverage of data provided by the suppliers is not enough and Messho as a platform runs only on data!

In this area, Meesho relies heavily on deep learning techniques to get insights from the images suppliers share and use that as another source of data to do the demand supply matching. Since the picture tells a lot about the product, the company uses this image to extract data like the color of the product, length of a Kurti or a Ghagra, and the print on it.

“This helps us increase the coverage of data from 50 to 100%. We then use this data to feed in the personalisation model to deliver the correct options of product to the consumers,” he concluded.

Join the community of 2M+ industry professionals

Subscribe to our newsletter to get latest insights & analysis.

Download ETCIO App

• Get Realtime updates
• Save your favourite articles

Scan to download App