Meesho to double down on AI/ML investments amid profitability push

Amid the ever-increasing integration of Artificial Intelligence (AI) and Machine Learning (ML) across sectors and enterprises, homegrown e-commerce marketplace Meesho plans to double down on its investments in AI/ML as the company aims to achieve profitability in FY24.

“AI/ML has been a big part of our product evolution in the last two or three years. And we are going to continue investing in it quite a bit going forward,” Kirti Varun Avasarala, Chief Product Officer, Meesho, told Business Standard.

“Investments in AI are going to be the biggest focus area for us from a technology standpoint. We plan on leveraging the latest advancements in the sector, be it generative AI, Chat GPT, or any other emerging technology, to improve the customer experience,” he says.

The company has been betting big on data science and technology-driven use cases over the past three and a half years.

“We have a sizable AI team of 50-55 people, and have used AI/ML across the board in our demand, supply, and fulfillment channels,” says Avasarala.

Current Offerings

On the supply side, Meesho has used AI/ML for cataloguing– a process of finding out what kind of attributes a listed item should reflect to customers.

“To make it easier on our sellers, we do not ask them to mention many details about items they list. This process of determining key attributes is carried out using data science, on a very large scale, and is quick and fully automated,” Avasarala says.

This, he explains, is carried out using ‘taxonomy tagging’ – assigning relevant attribute values to a product, which automates the process by leveraging AI. This eliminates supplier and agent tagging stages and has resulted in a 30 per cent reduction in manual efforts for Meesho. Moreover, the cataloguing time has come down by approximately 99.6 per cent from two days to ten minutes, while overall costs have been reduced by 45 per cent.

“We are also figuring out use cases on the seller side, where we can aid sellers in determining the right pricing for their products,” he adds.

A key area where we have used AI/ML at the consumer end has been for personalization. The firm offers customers various avenues for a personalised experience via search mechanisms, vernacular aid, etc.

More than 50 per cent of Meesho’s users are first-time e-commerce customers. “For these customers, a key area, from a data science perspective, that we have invested in is understanding what they want to buy when they come onto the Meesho app for the first time,” Avasarala says.

The platform, therefore, maps the demographic and vernacular data of existing customers and applies it to new customers to predict their preferences.

Meesho also offers generative AI use cases such as AI chat and voicebots, an area where investments are expected to increase.

On the fulfillment side, Meesho has used AI/ML to reduce cases of RTO (return to origin), where customers end up returning their orders. This, Avasarala says, happens more frequently in tier 2 and beyond regions where addresses aren’t as well defined.

“We use AI/ML to refine these addresses and reduce errors. Furthermore, we are using AI/ML to reduce cases of fraud as well,” he says.

The Secret to Tier 2+ Success

The firm, which is a frontrunner in tier 2 and beyond markets, has found success in these regions on the back of several factors.

For instance, as many as 65 per cent of products on Meesho are cheaper by 20-30 per cent versus other platforms, claims Avasarala. The factors driving this lowest pricing, he says, are competition among sellers, zero seller commission, and outsourced logistics that ensure there is no cost drag due to underutilisation of logistics capacity.

Meesho also has the lightest e-commerce app on Android in India with a compressed size of 13.6 MB, which is an advantage in areas with poor internet service.

The platform is, however, incrementally seeing traction from customers from metro/tier 1 markets across income classes, who are looking for affordable shopping alternatives.

During its early years, Meesho primarily focussed on apparel, and has now diversified into other lifestyle categories to become a ‘pure play horizontal marketplace’. As of Q4CY22, about 50 per cent of Meesho's GMV was from non-apparel categories as against just 30 per cent in Q4CY20, and this trend continues.

The Sequoia-backed firm logged 120 million average monthly active users (MAUs) last year, adding 100 million MAUs over the last two years. Meanwhile, the company’s monthly transacting user count rose 26-fold over CY20-22, resulting in a nine-fold increase in GMV. The company clocked 1 billion orders, up 2.2-fold year on year, and 140 million transacting users in 2022.

Meesho recently became the world’s “fastest shopping app” to cross 500 million cumulative downloads. It also became India’s fastest e-commerce platform to onboard 1 million sellers, with more than 600,000 small businesses signing up in the last year. The company achieved this feat within eight years of inception.

Cost Cutting

Its AI/ML-driven use cases have helped Meesho in cutting costs as it looks to improve its margins and achieve profitability.

“The ROI is very high on these investments, purely because there are no variable costs associated with data science,” says Avasarala.  

The company reduced its monthly cash burn by over 85 per cent to $5 million currently from around $40 million in early 2022, and is now looking to trim its annual revenue growth target to 40 per cent from the 100-plus per cent earlier.

In FY22, Meesho saw its revenues jump 4.5 times year-on-year to Rs 3,232 crore, while its losses rose even more steeply by 7.5 times to Rs 3,247 crore.  

Future Investments

Going forward, the company plans to double down on further AI/ML integration across its marketplace.

“One of the newer areas where we're investing in is, for instance, voice search. Customers who are not comfortable typing can use the voice search feature to search for products in their vernacular languages. We are working to enable a much more advanced version of this feature in the future with improved accuracy,” Avasarala says.

Meesho is looking to invest in improving user recommendations using AI as well. “If a customer expresses interest in a product, we will use data science to show them related products,” says Avasarala.

“There are some ongoing discussions about augmented reality, where we do see a lot of potential,” he says, adding, “We are trying to assess where we can use this technology to solve problems. But we are still in the ideation stage,”  Avasarala added.