Coffee Data Science
Determining Peak Flavor of Coffee using Bean Density
Reordering someone else’s data
A few weeks ago, I was pointed to Richard Mayston’s website where he explored over many months coffee bean density. He explored using bean density to determine the best date to drink coffee as well as the time to wait to start using a new bag of beans.
He had optimized other parameters based on bean density. There is a potential that a single measurement could give a great starting point for grind as well as other shot parameters such as temperature, pressure, dose, and output ratio. This potential would be a gift for coffee shops to avoid the waste of dialing in.
So I took a peek at his data, and I felt compelled to reorder it so I could see it better.
Background
Richard’s main idea is that if a coffee roaster printed bean density on their bag, it would eliminate some of the variables between how people roast. It would also help with roast to roast variations. Then that bean density could inform when to start using beans so that you could waste less.
His current website tells you how to cheaply measure the density, and then if you input that into his calculator, it will give you the parameters he used and the date range.
He evaluated multiple bean densities by having the same green beans roasted to different levels. Then he tried it as an espresso, espresso with bypass water (added hot water), and then espresso with milk.
The Data
I didn’t like having to search a bean’s density, and I wanted to look at general trends. He build a model to get more granularity out of his data, so I spent some time scraping the data for density intervals of 0.01. The values in his data go from 0.304 to 0.553 g/ml.
First, I looked at a few trends. He modified a few parameters based on density to give an optimal shot. The only criticism I have is that he didn’t measure Total Dissolved Solids (TDS) and Extraction Yield (EY) throughout the tests, so this is a subjective taste test. He tried to mitigate some of the subjectiveness using blind tasting by people other than himself.
We can take the day ranges for each density, and we can plot them. We can look at passible, good, very good, and best, or just very good and best. The trend is clear, and it’s interesting to see the darker roasts go bad very quickly.
I asked him if he recorded grind setting per density, and he graciously sent me that data. He used a Flair Pro 2 (45.5mm basket), and his grinder was a Kinu M47. According to his data, there is a solid trend in best grind per density.
Recommendations:
- Pay attention to roast dates especially for darker roasts.
- Use this as a guide and discover what you like best for yourself.
- Track bean density and grind setting for your grinder, and over time, you should be able to have a better grind setting to start with.
- Buy coffee in a way that allows you to get maximum flavor.
I’ve started to record bean density for my home roasts. So far, his data aligns well with my data and recommendation for medium roasts which is that they peak in flavor and extraction 3 to 5 weeks after being roasted. Hopefully, over the next year, I can get better data on the topic by including bean density in my data table.
If you like, follow me on Twitter and YouTube 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.
Further readings of mine:
Collection of Espresso Articles
A Collection of Work and School Stories
Measuring Coffee Grind Particle Distribution using Image Processing
A Summary of the Staccato Lifestyle
Measuring Coffee Grind Distribution
Espresso Baskets and Related Topics