Hi everyone, new member here I hope it’s OK to revive this topic.
I’d like to add this article a context to the TikTok case study: [Seeing Like an Algorithm — Remains of the Day] .
The whole 3 articles serie is excellent, but this article is the main one I think.
The keypoints are (with my commentary):
- Shorten the onboarding has been key to TikTok success. For that they decoupled the social graph from the user feed and used their paid content in the feed.
- => Remember that the paid content was crafted to please to specific population. This means that this content help the sorting algorithm decide what the user likes and which niche they are into. This implies a deep understanding of the tech needs from the marketing departement.
- The whole app’s UI is built around for give effective signals on user specific preference (ie: showing one video at a time allow the user to “vote” for a video by staying on it or moving elsewhere).
- => It’s probably needs a new app to have a UI that give sufficient signal to the sorting algorithm
- Tiktok spent half of its engineering budget on their algorithm needs and infrastructure.
- => don’t underestimate the amount of effort and expertise needed to reproduce this feat. It needs a solid technical leadership.
You can say that every internet B2C app needs a deep integration of the tech in the company functions to be able to function at scale, and that’s true, but I think it’s interesting to underline what they succeeded to do here.