Case Studies

"The Most Python" Report

Analyzing 200K Jupyter notebooks and 2M+ Stack Overflow posts to map how people actually write Python.

OVERVIEW

What 200,000 Notebooks Reveal About Real Python

An analysis of 200,000 IPython/Jupyter notebooks and more than 2 million Stack Overflow posts to find Python's most common code, libraries, and questions.

Instead of opinion about what Python developers should use, the report measures what they actually use, drawn from public code and public questions at scale.

The work became a widely read dashboard and write-up on how Python is really written in practice.

The Challenge

Turning two large, messy public corpora into a clean picture meant solving several problems at once:

  • Collecting and parsing 200K notebooks of varying quality

  • Processing 2M+ Stack Overflow posts

  • Extracting libraries, functions, and patterns from raw code

  • Aggregating it all into comparable, rankable signals

Links

APPROACH

The Approach

I built an analysis pipeline that:

  • Parsed notebook and post content into structured records

  • Extracted imports, library usage, and code patterns

  • Counted and ranked usage across both corpora

  • Compared what people write against what they ask about

  • Published the results as an interactive dashboard

Core architecture:

  • Python for parsing and analysis

  • BigQuery for storage and large-scale aggregation

  • SQL for ranking and comparison queries

  • Looker Studio for the public dashboard

Technical Stack

  • Python

  • BigQuery

  • SQL

  • Looker Studio

  • Data Visualization

RESULTS

Key Outcomes

  • Analyzed 200,000+ Jupyter notebooks

  • Processed 2,000,000+ Stack Overflow posts

  • Ranked Python's most-used libraries, functions, and patterns

  • Published a public dashboard and write-up

  • Built a repeatable method for measuring real-world language usage

Why This Matters

Technology choices are often driven by hype rather than evidence. This report showed a different approach: measure behavior at scale and let the data describe what is actually common. The same method (analyze large public corpora to map real usage) applies well beyond Python, to any ecosystem where adoption is debated more than it is measured.

Looking Forward

Possible extensions:

  • Tracking library usage trends over time

  • Comparing usage across languages and ecosystems

  • Linking common questions to common errors

  • Refreshing the analysis as the corpora grow

Interested in building something similar?

We help organizations design scalable AI and analytics infrastructure for:

  • forecasting

  • growth analytics

  • operational automation

  • machine learning platforms

  • agentic AI workflows

  • cloud-native data systems

Let’s build systems that turn data into operational leverage.