Close Menu
    I home roast coffee and could never keep track of origin flavor profiles so I mapped 32 of them [OC]
    Data

    I home roast coffee and could never keep track of origin flavor profiles so I mapped 32 of them [OC]


    I've been home roasting coffee(using a Fresh Roast SR800) for ~2 years and kept running into the same problem: I'd read tasting notes like "bright, floral, citrusy" for one Ethiopian and "earthy, winey, fruit-forward" for another, and eventually it all just blurred together. I wanted something visual that actually let me compare origins side by side.

    So I mapped 32 single-origin coffees across six flavor dimensions(Fruity, Floral, Sweet, Nutty, Spicy, Earthy) and built an interactive site around the data. Scores are calibrated to specialty-grade, light-to-medium roast for each origin.

    Some things you can do with it:

    • Compare any two origins head-to-head with an overlaid radar chart
    • Explore a PCA scatter plot that clusters origins by flavor similarity
    • Filter by brew method or roast level
    • Pick flavor tags you like and get back matching origins ranked by fit

    Just got into espresso last month and the Compare view has been super useful for figuring out which origins might work well together as a blend.

    Site is free, no ads, no signup. Source code is on GitHub if you want to poke around.

    tastingterrain.com

    Posted by neilkod

    View 4 Comments

    4 Comments

    1. neilkod on

      Scores and tasting notes are synthesized from Cup of Excellence auction reports, Coffee Review, Sweet Maria’s green coffee library, the World Coffee Research Variety Catalog, and regional coffee board data. Each origin is scored 1-10 across six dimensions at specialty-grade, light-to-medium roast.

      All 32 origins live in a single JS file in the repo: [github.com/neilkod/tastingterrain.com](https://github.com/neilkod/tastingterrain.com)

      **Tech stack:**
      – React 19 + Vite 8
      – No charting library, all radar charts and the PCA scatter plot are hand-drawn SVG
      – No CSS framework, all inline styles
      – PCA implemented in ~30 lines of vanilla JS
      – No backend, no database, no API, just a static site deployed on Vercel

    3. foltranm on

      as a Brazilian who loves nutty coffee, I’m very pleased with the coincidence of this graph 🙂

    4. coofffeeee on

      I kinda love this. Gonna drive in a bit more when I get more time. Will likely be back with feedback and questions. Thanks!!