Coffee is a beverage with a rich history and an incredible diversity of styles. Recently, I analyzed a dataset of over 7,000 coffee reviews to visualize how different roast levels compare in terms of tasting notes. The data source and Jupyter notebook are linked at the bottom of this post.
The data is from thousands of coffee reviews from coffeereview.com, so it is a bit noisy. Some data was clearly labeled (e.g. roast level), but the tasting notes were contained in the free-form review notes. I used ChatGPT to process the review text and extract the tasting notes, which allowed me to categorize them into common flavor profiles.
Key takeaways:
- The most common tasting notes are floral, fruity, nutty, and sweet notes.
- Darker roasts are less likely to have fruity or floral notes, and more likely to have burnt, and papery / musty notes.
- Darker roasts received lower ratings, on average.
- The most frequently reviewed coffees are light to medium roasts.
Download high-res image
Sources
The figure was made using Python’s Plotly library and Figma.
Data Source:
Jupyter Notebook: