In 2019, the Indian government banned TikTok. In the same year, home-grown short video platforms boomed in India. Chingari, Mitron, Josh, Taka Tak, Moj, Roposo and many other platforms came into existence in the subsequent years to fill the gap. 

Three years later, these platforms are still struggling to get a foothold in the short video space and face stiff competition from newly-launched Instagram’s Reels feature. Experts say that these short video platforms have missed the point of TikTok becoming TikTok. They say that it wasn’t the short videos idea of TikTok which revolutionised the video industry but its recommendation engine.  

TikTok knows you better than you know yourself. It understands the unique taste of every individual. It has a secret algorithm which quickly figures out users’ behaviour and recommends videos of what they like to watch. 

Here’s how: In 2012, China-based Bytedance launched ‘Toutiao’, a news aggregator, which was a major hit in China with its recommendation-based news system. 

Four years later, in 2016, Bytedance announced ‘TikTok’, which changed the video consumption process forever. 

The magnanimity of TikTok is reflected through its numbers. In just five years, it has acquired about 1.2 billion monthly active users (as per Q4 2021) and is estimated to reach 1.8 billion users by the end of 2022. The short video app has been downloaded over 3 billion times. 

While TikTok’s popularity continues to rise elsewhere, Indian short video apps are trying to understand their users’ behaviour.  

How homegrown apps algorithm works 

Chingari received more than 15 million user downloads following the TikTok ban. It has over 130 million users and over five million daily active users watching videos in 15 languages. 

The company claims that it is continuously developing its recommendation engine to build regular engagement with the user base. The objective of recommendation for both the systems is to generate content which is more relevant for the user.

In an exclusive interaction with AIM, CEO and cofounder Sumit Ghosh, Chingari, elaborated on their recommendation system: “We use a combination of multiple recommendation systems to generate a diverse set of videos in the feed.”

He further added that there are two main systems—popularity driven feed and feed generated,  which is trained on user interactions (watches, likes). The ML model output will be personalised for each user based on their past interactions while the popularity feed will generate fresh popular videos every few hours based on what’s currently tracking well.

Sumit Ghosh, CEO and cofounder at Chingari

Besides, the platform is utilising machine learning to personalise notifications for users. “We are moving towards more hyper personalisation with our new real time feed and will focus primarily on what the user interest is at a session level,” he added.

Ghosh further shared the challenges during the hyper-personalisation in the recommendations: “We need to collect data for personalising content for each user, and customising their user experience is one of the biggest challenges we face. Secondly, feature engineering on the data collected is an extensive task that requires time to convert them into consistent, good quality actionable insights using machine learning models.”

Scaling systems to handle real time user interests to provide content becomes very crucial for the success of any recommendation system when it comes to hyper personalisation. Ghosh also said that applying hyper-personalisation for cold start users who are completely new to the platform poses a significant challenge. 

Taking the path less travelled 

Tiki is another short-video platform that has stolen the thunder of  mainstream platforms. 

Ian Goh, Co-founder & CEO at Tiki spoke exclusively to AIM, “Our recommendation engine is designed to serve real talented creators, not the other way around. Many platforms are using their engine to take advantage of their creators, but that’s not what we do.”

Ian Goh, Co-founder & CEO at Tiki

Ian claims that the platform has a very active user base in cities like Uttar Pradesh, Maharashtra, Rajasthan, Karnataka, Gujarat, and Bihar.

“[About] 60% of our traffic comes from Tier2 & Tier3 cities across India. With the rise of internet users in India, many Tier 2,3 cities users have come online. Most social platforms may be too saturated for them and it is not easy to grow their followers now.”

Along with such home-grown social media apps, tech giants are also working to better understand users’ behaviour.

Big tech’s obsession with TikTok

Major tech giants with their social media platforms have been trying overhaul with features similar to TikTok. For the first time in its history of 18 years, Meta reported a loss in the count of its daily users earlier this year. Meta’s CEO Mark Zuckerberg was quick to blame TikTok for user stagnation in the company’s Q1 2022 earnings call. 

The launch of ‘Reels’—a short-video product—that made up more than 20% of the time users spent on Instagram was one such imitation of the China-based platform.

Also read, ‘Make Instagram Instagram Again. Stop trying to be TikTok’

Meta also introduced cross-posting from Instagram to one of its other platforms—Facebook, enabling creators to increase engagement. YouTube introduced a new feature—‘Shorts’—that is also eligible for monetisation, where creators would keep 45% of the revenue generated from their viewership.

Further, Amazon is also testing a feature on its platform to show users photos and videos with products for shoppers inspired by TikTok’s style of interface. 

The million-dollar TikTok code

Interestingly, TikTok started off as a streaming service used by millennials for lip-syncing but evolved into users generating millions of dollars using their personal content. What’s more important is tapping on to the right audience, which includes providing the right recommendations for users to love what they see.  

TikTok Boom’s author Chris Stokel-Walker says, “One person at TikTok in charge of tracking what goes viral and why told me that there’s no recipe for it, there’s no magic formula. The employee even admitted that it’s a question that they don’t think even the algo team has the answer to. It’s just so sophisticated.”