Why Is Tik-Tok So Addictive? A Data Perspective.

Published 22 Jul 2021

By Saumil Talati

You wake up.10:00 am. The sound of your mum vacuuming outside and the constant and repeated yell of the kookaburra bleed your ears. You turn to your side and grab your phone.

Inside you know you might regret this, but nevertheless you open the app. Tik-tok. You’ve only been on the app for 5 minutes, yet as you look at the clock on the wall in front of you – it reads 11:00 am.

Why is Tik-Tok so addicting? What is it about the platform that just makes it so addicting?

What’s the answer? Data.

The key to Tik-Tok’s meteoric success, its gain of 682 million users just last year has been its recommendation engine. A recommendation engine is essentially a type of information filtering system that seeks to predict the rating a user would give it. The specific type that Tik-Tok uses is a collaborative filtering method. This relies on the assumption that a person will like similar things that they have liked in the past (if you liked cute puppies in the past, you still like them now). Then by matching users with similar rating history, they can generate the recommendations.

Recommedation engine flowchart Figure 1: Workflow of Tiktok.

One key advantage that Tik-tok has, compared to other social medias such as Youtube, and the reason their recommendations are so good is the number of data points. Let’s take Youtube for example. An average YouTube session lasts for 40 mins, with an average of 7 videos. An average TikTok session lasts 10 minutes, with an average of 35 videos that are watched.

Since the Tik-Toks are so short in duration (60s max), the algorithm has much more datasets to work with, and thus is more effective in selecting videos which you will like. It simply can trial and error much more freely.

Ok, that’s great, but what kind of data points are collected?

The obvious ones for a start are:

  • Hashtags
  • Likes
  • Comments
  • Followers
  • Age, gender, location

Tiktok’s biggest indicator however, known as ‘strong indicators’ is if the user finishes the video from the beginning to the end.

So, that’s all great. But how can you use this to your advantage? How can you, using data, be able to have the greatest chance of a viral Tik-tok video?

Here are 3 tips:

  1. Make the videos short – The goal is to make the user finish your video! 15-20 seconds is a great length
  2. Post when your user base is Online – go to your analytics and check the country in which the majority of your audience is from. Then only post when they are awake!
  3. Don’t delete your old videos – We know it can be discouraging having videos with only a few hundred views. In-fact they can become viral due to ‘batching’. This is where Tik-Tok will expose your videos to a new demographic periodically.

So finally, there we have it! As you can see, data is everywhere. Hopefully this article was interesting! Make sure to try the tips above so your next Tik-tok is Charlie Level.

Some Additional Information for Curious minds :

Have you sometimes wondered how smaller creators can get millions of views? Yet, with other forms of social media such as Youtube, only big creators tend to get large numbers of views. This again can be answered by the recommendation engine that Tik-tok uses. Although Youtube also has a recommendation engine, the subscription model it uses has a major influence on that. However with Tik-tok, although more followed accounts (Charlie, Addison) are more likely to get more views, neither follower count nor previous high performing videos are taken into account within the recommendation system.


Towards Data Science: https://towardsdatascience.com/why-tiktok-made-its-user-so-obsessive-the-ai-algorithm-that-got-you-hooked-7895bb1ab423

Tinuiti: https://tinuiti.com/blog/paid-social/tiktok-algorithm/

Tags: Data Science Applications of Data Science Machine Learning, Deep Learning and Neural Networks