How to Design an Algorithm-Friendly App: Insights from TikTok and Beyond

How to Design an Algorithm-Friendly App: Insights from TikTok and Beyond

In today's digital landscape, machine learning algorithms are becoming increasingly integral to the success of companies across all industries. TikTok, the popular short video platform, is a prime example of how an algorithm can be a game-changer. While many experts doubt that TikTok has made groundbreaking advancements in machine learning recommendations algorithms, it is undeniable that the effectiveness of an algorithm depends on the data it is trained on. TikTok's design holds the key to its algorithm's success.

One of the secrets behind TikTok's algorithm is its closed loop of feedback. This feedback loop inspires and enables the creation and viewing of videos that the algorithm can learn from. But to truly serve your users, you must first serve the algorithm. Every action you take on TikTok, from the moment a video starts playing, sends a signal to the algorithm about your sentiment towards that video. Even before the video reaches your phone, a human on TikTok's operations team has watched it and added relevant tags or labels. Vision AI is also invoked during the creation process, with camera filters designed to track human faces, hands, or gestures.

The design of TikTok's algorithm-friendly app goes beyond the video creation process. It challenges the traditional infinite scrolling feed used by many social networks like Facebook, Twitter, and Instagram. Rather than displaying one story at a time, TikTok shows multiple videos on the screen simultaneously. This approach poses a challenge for the algorithm to accurately judge sentiment, as it cannot determine which video captures your attention as you scroll. The current design of social networks prioritizes lower friction scanning for users at the expense of accurately capturing negative signals.

Moreover, content derived solely from a user's social graph can lead to a mismatch between their true interests and those of their connections. This can cause the algorithm to drift away from the user's genuine preferences. To address this issue, algorithm-friendly design takes a different approach. It aims to serve the user's interests by understanding their needs and preferences, rather than solely focusing on minimizing friction. The goal is to help users achieve their desired outcomes.

In today's software era, true competitive advantages are becoming increasingly illusory. Features and UI designs can be easily copied by competitors, rendering them ineffective. The true magic lies in how every element of an app's design and processes connect with each other to create a dataset that trains the algorithm to achieve peak performance. This requires a deep understanding of how the flywheel works and a commitment to align every element and process towards a single purpose and goal.