Coffee Grounds Freshness for Espresso

Exploring time differences between grinding and brewing

Published in

Towards Data Science

7 min readSep 4, 2020

How long could you wait to brew espresso after grinding? Ask a barista, any barista, and I suspect their answer is immediately or within a few minutes. The common knowledge on coffee grounds freshness is that after 30 minutes or so, coffee grounds go stale. Even if stored, the common knowledge is that coffee grounds still go stale (the length of freshness is not well defined). Is that always true?

My experience with the staccato shot has led me to believe that the common thinking on coffee grounds freshness is not correct. I have looked for supporting data, and there is not much, probably because it seems so obvious. So I will share my experience and my data which suggests that if stored in an airtight container, coffee grounds can be used for at least a few days without losing flavor. The past few months, I’ve been storing my the refrigerator for my spicy shots, but I have also looked back at coffee grounds before I refrigerated them regularly.

Previous Experience

For awhile, I was grinding fresh, but that changed when I started to develop the staccato espresso shot. Grinding and sifting for every shot was tedious, so I would grind enough coffee for a few shots, and I would sift and store the coffee in air tight containers.

I found such a benefit to the containers that I would grind coffee and store it in them even if I wasn’t going to sift it. To make my process more efficient, I would grind in the morning, and an hour later, I would sift the grounds on my way to work. The time between grind and brew started extending.

I also am lazy when it comes to certain things, and I have on occasion left grinds in sealed containers for multiple days. I didn’t notice much of a degradation in taste.

Data Collection

I started recording grind date and time only when I started recording bean temperature at grind. I was doing experiments with hot and cold beans at grind, and I figured I would keep track of it. I typically would grind four shots at a time, and it would take me 2 to 4 days depending on how much time I had to drink fancy coffee.

I started looking at that data to see if I could tell me if there was any change in taste or extraction yield. When I didn’t see a change (as seen further down), I did a few longer experiments where I would grind 8 shots worth of coffee to drink over a week. I also went back to previous data and labeled estimated grind dates based on when the grounds were sifted, which I recorded separately.

Most of these shots were using home roasted beans that were 1 to 1:30 minutes past the first crack. I ground the beans on either a Rok grinder or a Lime grinder, and this data is from shots pulled on a Kim Express. Additionally, if one thinks having a better grinder is more important, I would refer them to this work where I used a blade grinder and a sifter to make a good shot of espresso which suggests a sifter equalizes the field of grinders.

Metrics of Performance

I used two metrics for evaluating the differences between shots: 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 and affected the final score.

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

Data Findings

First, looking at the data, I plotted all the data, and I didn’t notice anything. There was definitely variation over the days in taste (Final Score) and EY, but I didn’t see any patterns. I will note that I stored most of these grounds in the refrigerator due to spicy grinding, but I haven’t noticed a taste degradation for grounds stored at room temperature for a day or two.

So I broke it down by roast, and I shorted the time window to 3 days maximum. Again, each roast seemed not to degrade over time. There was variations within the day related to changes in some experiments I was running, but there was not a trend.

I decided to simply the data. I cut it in half because half of the data had a grind age of less than 1 day and 4 hours. I wanted to do within a day and greater than a day, but it wasn’t an even cut. In this split, there were some differences, but none of the distribution differences were statistically significant.

I typically would grind enough for four 18g shots, and I wanted to test the grinds over a longer window than previous. I did this with two roasts, and I saw variations, but not trends. The grounds were also stored in the refrigerator which could have contributed to their taste staying fresh.

Shot Time

Shot time also seemed unaffected by grind age. Across the different roasts, the shot time stays pretty consistent, and I didn’t see any trends.

More Data

I pulled 359 shots that I have grind time and time of the shot, but with the wealth of data comes the difficulty of trying to make sense of it. So I first plotted it as a scatter plot. No trend is apparent.

Then, I plotted box plots, and it seems the Final Score or taste improves, but I suspect that improvement is due to improving shot parameters. Also, the distributions are very wide due to the shots being across different roasts and techniques. The is a slight shift in median score, the distribution still stays pretty similar.

To extract learning, I normalized (Z-norm) the data by roast. For each roast, I calculated the mean and standard deviation (std). Then, for each metric, per each roast, I made the normalized data by this equation: X_norm = (X – Roast_mean)/Roast_std. If there was a trend due to grind age, it should be clear as a change of the normalized distribution.

As a scatter plot, there doesn’t seem to be much of a change, so let’s look at box plots below. For Regular shots, Final Score stays about the same while EY increases a little bit.

For Staccato shots, there is not a trend for any of the metrics. 0 in this case means the mean score while positive is higher than the mean, and negative is lower than th emean.

After a suspicion that grind age didn’t matter as much as has been suggested, I looked through as much data as I could find. I didn’t find any trends positive or negative between grind age and multiple metrics (Final Score (taste), EY, or shot time). The main caveat of ground coffee losing taste is most likely based on leaving it exposed to air for a period of time, but at least from my experience, grind more less often has not had a negative impact to my espresso experience.

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.