Cold Start Problem: New Users & Items

Learn why platforms struggle with new users who have no history, and how they solve it

Part of How Netflix, Spotify, TikTok Know You

4 blocks0 nested pages

Last updated Oct 29, 2025. Clone to remix or explore the blocks below.

ba6bc4a7...

4c9b0274...

content

TikTok's Genius: Why Those First Videos Feel Random

Educational content slides

You download TikTok. Open it. First hour feels chaotic:

Cooking video (watch 2 seconds, skip)

Prank video (watch 3 seconds, skip)

Anime edit (watch 47 seconds, like, comment)

Dance video (watch 8 seconds, skip)

Anime meme (watch full video, share)

Within 20 minutes: TikTok knows you better than Instagram after 2 years

The secret: Aggressive exploration with rapid feedback. Each skip/watch/like is a data point. Short videos = fast learning.

This solves the cold start problem in real-time.

a469bc5c...

quiz

Test your understanding with this quiz.

0 / 3

Why does Spotify ask new users to pick 3 favorite artists during signup?

69880f10...

feedback

Complete this exercise and get AI-powered feedback.

Cold start is every recommendation system's first challenge. How you solve it determines whether users stay or leave.

Think about: How to gather data quickly without annoying users, when to explore vs exploit, how to balance learning and satisfaction.

This reveals: The fundamental tension between learning and serving in recommendation systems.

Design a cold start onboarding for a new app (books, recipes, study resources, events, etc.)

How would you balance learning about users (exploration) vs giving good recommendations immediately (exploitation)?

Progress5 of 6