HyperLogLog: The Algorithm That Counts Billions with Just 12 KB of Memory

Anurag Goel

11 min read

·

Jun 27, 2026

--

Listen

Share

TL;DR: HyperLogLog is a probabilistic algorithm that estimates the number of unique items in a massive dataset using a tiny fraction of the memory a traditional approach would require — with less than 2% error. Used by Redis, Google Analytics, and Mixpanel, it’s one of the most elegant engineering solutions in Big Data.

Press enter or click to view image in full size

Table of Contents

1. The Problem: Counting Unique Things at Scale
2. The Intuition: Finger Counting with Probability
3. Evolution of the Algorithm

• Flajolet-Martin
• LogLog
• SuperLogLog
• HyperLogLog

1. How HyperLogLog Works (Step by Step)
2. Go Implementation with Redis
3. Real-World Use Cases
4. Pros and Cons
5. When Should You Use HyperLogLog?
6. Conclusion

The Problem: Counting Unique Things at Scale

Imagine you run a streaming platform with 500 million users. Every day, your product manager asks: “How many unique users played this song today?”