We have Raised US$502 Mn to Build India’s Largest AI-Powered Content Ecosystem
It has been an exciting 6 years for us building Moj & ShareChat and today is an important milestone in our journey.
We have raised US$502 mn at a US$2.1 bn valuation from Tiger Global, Lightspeed, Twitter, Snap, and other investors.
It feels magical mindsight to see how a simple insight on WhatsApp group usage in India could lead to such an impactful journey.
I and my cofounders Farid & Bhanu were in IITK when we participated in a hackathon in 2012 and met each other.
We built 14 products while being in college — and failed 14 times to get it off the ground.
But in Nov 2014 we discovered an interesting phenomenon — the interest-based WhatsApp groups in India.
We discovered that Indian users were hungry for finding content in their own language. They were sharing their phone numbers publicly on FB in the hope to get added to an interest-based local language group on WhatsApp.
We tried creating/joining 10–15 such groups and saw users using it for all kinds of purposes — getting news, joining celeb fan clubs, finding new jokes, shayaris, devotional content, etc.
With this insight — we released our first version of ShareChat in Jan 2015 — a simple app to find trending content in Indian languages with a baked-in viral loop — one-tap share to WhatsApp. It worked! 🔥
Over the years, we have seen ShareChat evolve tremendously and become India’s largest regional social media platform with 16Cr MAUs serving 15 Indian languages.
The content diversity on the platform has increased dramatically — you can watch a popular humour video while also getting an Ayurvedic doctor sharing a home remedy.
We had seen short videos growing in popularity on ShareChat and externally on other platforms in the last 2 years and understood that there is a large appetite for this format.
When we saw a large vacuum emerge on June 29 with a lot of short-video apps exiting the market — we knew this opportunity was for us. There were millions of short-video creators already trained for creating that content supply. The real game was therefore going to be on building the most relevant feed for the user and we were the only Indian company that had built a world-class feed recommendation system for short-form content.
We built the app in 30 hours. Scaled it to 12Cr MAUs in 9 months. 🔥
AI-powered content feed
On both ShareChat & Moj, the content consumption is driven through an AI-powered content feed. Social actions like Follow are a part of the product but have limited impact on our content recommendation.
We believe social graphs are not a necessity to rank and personalize public content.
It is slow and inefficient to ask our users to follow accounts before we can serve them a good feed.
A very simplistic summary of how we do this — We focus on how to best match a new content with a minimum set of relevant audiences and keep expanding the exposure of that content in the neighborhood of users who engage with it.
Historically, it has been necessary for any user to add an adequate number of relevant connections on any social media platform in order to derive value from the product. Now, the motivation to connect — Follow or Friend other users is not always clear. You may readily connect with members on social platforms if you know them, either because they are a celebrity or they are an acquaintance but you don’t do otherwise. As a result, growing connections can be slow and we observe that many users tend to churn before they reach a critical number of connections. An AI-first feed seeks to delink your user experience from the activity of your connections and the user can get feed recommendations from a much larger pool of content from across the platform.
So, not only do we get to a richer and more personalized feed but also we do so more efficiently.
Another important application area of AI lies in the automated understanding of semantics and the quality of content. We develop multimodal machine learning models, which can process visual, audio, and text data to make machines interpret content in the same way as humans do. This is crucial for a couple of reasons. First, we use these representations to effectively match a post to the right set of users, especially in the early part of a post’s life cycle. Second, we apply these models to address different trust & safety concerns on the platform e.g., removing NSFW, baitish and spammy posts.
We have seen over years how the quality and speed of personalization is so highly correlated with the user’s long-term retention and therefore our focus on the AI-first approach in Moj has led to us having 50% higher retention and 2x the DAU compared to our closest competitor.
With a better-personalized feed, creators also get the most return for their effort, we are able to match their content to the largest set of relevant audiences even if they create content about a very niche topic.
We are at a significant inflection point in our company’s journey — as the internet penetration further deepens in India we are well-positioned to expand our ecosystem of products to 100 Crore+ MAUs cumulatively.
We have seen how large the short-video market is in China — with around 80% of the entire Internet population using one of the short-video products(Douyin, Kuaishou, etc) daily.
We are building a world-class distributed team in India & the US.