Coffee Data Science

Correlating Particles and Colors in Coffee to Extraction

An exploration into regular uses of particle measurement and color

5 min readOct 13, 2023

Coffee color and particle distributions are interesting, but for every shot of espresso? When I got the Omni, I wanted to see what ways the device could be useful for someone like me on a day to day basis.

So I started collecting data on every shot, and I wanted to see if there was anything with a good correlation to extraction yield (EY) and taste. This is part of a broad investigation to poke at what other data would be useful to collect during the coffee making process.

The DiFluid Omni can collect data on coffee color and particle distribution. This lead me to collect data on the following:

Bean Color
Grounds Color (Tamped)
Comparing Fluffy vs Tamped Grounds Color
Color of the puck post-shot (Top and Bottom)
Particle Distribution Buzzed (per design)
Particle Distribution Dusted (per my SOP on particle imaging)

Procedure

This is how I prepared each shot:

Measure the bean color
Grind the beans
Measure particle distribution using buzz function a sample
Measure particle distribution after dusting another sample
Prepare the puck but don’t tamp
Measure the grounds color (fluffy)
Tamp
Measure the grounds color
Pull the shot
Measure the color of the top of the puck
Carefully knock out the puck and measure the bottom

Comparing dusting to the buzzing function was slightly challenging. Either I dust after buzzing or use a different sample. Both have some measure of error. With buzzing, I am affecting the particles before dusting that might be different. Dusting particles negates the need to buzz. So I took different samples, which could have a bias. But this investigation was aimed as a starting point.

Metrics of Performance

I used two sets of metrics for evaluating the differences between techniques: Final Score and Coffee Extraction.

Final score is the average of a scorecard of 7 metrics (Sharp, Rich, Syrup, Sweet, Sour, Bitter, and Aftertaste). These scores were subjective, of course, but they were calibrated to my tastes and helped me improve my shots. There is some variation in the scores. My aim was to be consistent for each metric, but some times the granularity was difficult.

Total Dissolved Solids (TDS) is measured using a refractometer, and this number combined with the output weight of the shot and the input weight of the coffee is used to determine the percentage of coffee extracted into the cup, called Extraction Yield (EY).

Intensity Radius (IR) is defined as the radius from the origin on a control chart for TDS vs EY, so IR = sqrt( TDS² + EY²). This metric helps normalize shot performance across output yield or brew ratio.

Data

I collected data on 25 shots over 10 roasts.

I focused on correlation to reduce the number of variables to the more interesting ones. Correlation:

100% = Positive correlation
0% = No Correlation
-100% = Negative Correlation

Correlation to EY

Starting with extraction, the darker the bean color (lower number on Agtron scale), the higher the extraction. This has a strong correlation. This should be expected because darker roasts are easier to extract.

Fluffy grounds measurement was less correlated.

In terms of the spent pucks, the bottom had a better connection, but I was surprised how much the darker spots correlated to EY, just not the middle bins.

For particle distribution, there was a stronger correlation for the buzzed particles, which was curious. I’d like to see more. This seems to suggest a coarser grind was better.

Correlation to Taste

Taste was correlated only to one part of the bins, but not like extraction.

Taste had a stronger relationship to fluffy grounds, which was also interesting and unexpected.

For the post puck, there was less correlation than to EY.

Particle distribution didn’t seem well correlated.

I’m most surprised by the difference in correlation for fluffy vs tamped grounds. I wonder what the raw images look like. I would be curious to collect more on this variable because maybe the way the Omni takes an image, it detects some inherent piece of information related to taste.

If you like, follow me on Twitter, YouTube, and Instagram where I post videos of espresso shots on different machines and espresso related stuff. You can also find me on LinkedIn. You can also follow me on Medium and Subscribe.