Analytics Vidhya
Published in

Analytics Vidhya

Coffee Data Sheet

It was Christmas 2018, and all I wanted was a good shot of espresso from la Pavoni. I infrequently used the machine, and I was having trouble. After reading about preheating, I decided I needed to collect data on my shots to make sure the changes I made to the routine helped. Little did I know, the data sheet would become the most useful tool to making amazing espresso and expand to improved my coffee roasts.

I didn’t start tracking my shots because I thought it was some grand idea. It was more of a curiosity that turned into the single greatest tool I have to improve my espresso shots: data. One interesting retrospective aim is to be able to look back on historical data because some times, the value isn’t fully known until the data is available.

Honesty is the key in a data sheet. Without honesty, historical data means nothing. Without honesty, it is not possible to improve your espresso shot. Some times, I have wanted to score a shot higher because I thought it should have been higher or I was thinking about the average score, but I really want to stay honest in my assessments so that as time goes forward, I can hit the best shots of my life.

Eventually, the keys aims evolved over time to these five:

  1. Shot Improvements
  2. Fully Utilizing each Espresso Machine
  3. Bean Experiments
  4. Future unknown discoveries in method tweaks
  5. Palate Calibration

Input Variables

The input variables should be the most controllable and consistent across shots. Coffee weight is the easiest to control with a scale.

You could also measure water temperature, and 93C is widely viewed as the best temperature. I’ve done a small study on water temperature related to preheating the grouphead, but I haven’t specifically measured temperature per shot. It is an area that could be improved, but most of my machines, I figured out when to pull the shot before they overheat.

Other input variables are the coffee roast, date, time, weight per layer (for staccato), and I’ve added some other bits as optional like robusta added or the water level in the reservoir.

Process Variables

I have recorded the amount of time for pre-infusion, bloom, and extraction, but once I’ve tuned in my method, I often don’t time it. I have recorded many of my shots to help review how the shot went from the bottomless portafilter. This bit of data has also been essential for me to improve coffee distribution, dial in a grind, and balance staccato layers.

I have a few cells for notes. What slight variation I changed or some variable that I haven’t committed to tracking full time. Finally, there is a cell for what to do next time. Most of the time, I can remember what to do next, but just the process of documenting these notes has helped improve my ability to remember. Generally taking notes in class, at work, or at home has helped me tremendously.

Output Variables

There are two types of output variables: quantitative and qualitative. For quantitative metrics, I measure output weight and calculate the coffee grounds in to the coffee out. For awhile, I calibrated my espresso cup and measured percent solubles, but I haven’t in awhile. I have also taken photos for quite a few shots, and I could have tried to use image processing analysis to investigate the color differences, but I haven’t.

For qualitative metrics, I measure seven qualities: sharp, rich, syrup, sweet, sour, bitter, aftertaste. If I was a Q-grader, I could use those metrics, but alas, I am not. The first three measurements require explanation as the other four are self explanatory.

Sharp refers to the sharpness as the espresso first hits your tongue and almost how shocking it is to your tastes. I prefer not to cough as the result of drinking something so strong, but it is a sign of a sharpness I prefer. The one issue with this metric is that a smooth shot won’t be sharp, but that doesn’t mean it is a bad shot. So it has made me look at changes to which metrics I use or even going through the Q-grader certification.

Rich refers to the overall richness and complexity of the shot. I would prefer a shot of 1:1 ratio (input weight to output espresso weight) because I want to relive the first shot of espresso I had. My first shot tasted like I had melted a chocolate bar in my mouth; it completely took over my tongue.

Syrup refers to the texture, and I aim for what I recall my first shot to be. The espresso felt so dense on my tongue like syrup or melted chocolate.

When I’ve looked at the Q-grades, I have found some with similar descriptions as mine. Keep in mind though that Q-grading is not done on espresso. Here is a table with the closest Q-grades to mine.

Finally, I produce an average score that takes these seven metrics and combines them with a weighted value of the ratio of input to output. I’ve debated removing that ratio from my calculation because I’ve used a cup with measurement markers on it. This cup has allowed me to hit between a ratio of 0.8 and 1.1 every time.

My current cup

Calibration

I started with a small scale (0, 1, 2). I didn’t know how well I could distinguish between different levels, and I thought I would start small and expand as needed. Initially, I started grading on shots using a Hario grinder because I was on Christmas break. When I returned home to my Rok grinder and Kim Express, I quickly had to expand my scale from 0 to 4.

Then came the Staccato. I honestly didn’t predict a staccato shot would do so well. When I started experimenting, I started to see 5’s and 6’s in certain categories. As I got a better understanding of how the layers interacted, I was able to get each score category to go up. I migrated from two layers to three layers, and I saw the average score go from 4 to 7.

More recently, I have hit 9’s and slightly higher with some specialty roasts I have been tweaking. The data used for roasting requires a future discussion.

Every now and then, I check my calibration at a coffee shop, with shitty coffee, or on my superautomatic at home. While it is not always fun having a bad shot, it has certainly affirmed the lower end of my scale and has allowed me to see what those shots did well.

While in Italy, I scored a few shots, and I really noticed what I liked in those shots that I was not getting in mine. I determined I was missing out on using Robusta, and this drove me to experiment with tweaking the percentage of robusta to get the balance I so desired. Italian shots had better richness, sourness, and aftertaste, but they were not sweet. They were bitter, but not burnt. There is a difference …

Coffee Shops have the roast highlighted in Red. Noticed how the score patterns are quite different.

Multiple Machines

The fun part of this data collection has been working on multiple machines. I have a Kim Express at work, and that has been my baseline. I make my best shots on that machine because I’ve been using it for five years. Based on the past six months of data, I had 200 shots in six months, so 400 a year by five years would be 2,000 shots.

I bought an Enrico of Italy, and I keep it at my lab at work. It’s a good machine, but I have not used it nearly as often. The percent of accidents with that machine is much higher than the Kim (1 in 30 vs 2 in 2,000).

At home, I have an Oreo Giro, Flair, La Peppina, and a newly acquired Kim Express. I bought the new Kim Express because it needed a lot of love, and I wanted to have the same experience at work and at home.

I have done more shots on the Flair at home than any other machine. I almost gave up on the machine because I was having some trouble, but I eventually figured out some tricks, especially for staccato shots that allowed it to produce shots on par with the Kim.

Data

Let’s look at some data. First, let’s look at over time shot comparison. Usually, there is an ebb and flow that depends on a coffee bean.

Another way to view the data is the maximum score or shot per machine per roast. It shows a fun journey of how I used each machine and how my shots improved them.

At a higher level, I seem to have 2 to 4 shots a day, and I get close to 20 shots per roast. A roast is typically 350 green to 300 grams roasted, so 15g per shot means about 20 shots.

Average score across roasts over time changes as the beans age. I could especially tell from some beans that had a huge amount of crema at the beginning, which then diminished over time. Here is another way to look at shots per roast over time as well as understanding the shots per roast.

I can even look at shots post-roast date. We can normalize these scores by each roast and then combine them to have a deeper understanding of when is the best time to brew espresso. It seems like two weeks post-roast is a great peak, but I don’t have enough data post roast to understand when that flavor starts to go stale.

In my Staccato experimentation, I ended up with a lot of data. So here are a few ways I have cut the data to show how Staccato has improved shot quality over a regular shot. I have controlled for roast and machine.

1. Green (Staccato > Regular) , Red (Regular > Staccato). 2. Scatter Plot with red-line to show no difference

I could focus more on just two machines: Kim Express and Flair Espresso. In terms of solutions, it seemed to improve extraction to a point. I started to notice the lack of correlation, and I suspect extraction solubles is not the end of, be all metric for grading espresso quality.

While percent solubles is an interesting metric, it doesn’t predict taste (i.e. Score) well.

The God Shot

I used to think I could achieve some kind of “God Shot.” One shot that would be so amazing, it would ruin all other shots. I have hit this kind of shot many times, to the point that I think it is only dream, a fantasy that I’ll ever be satisfied with the taste of my shot. I don’t mean to brag; I mean to say I have hit a few levels of espresso goodness that I hadn’t achieved before, and then it became normal.

According to my data, I’ve made the best shot of all the previous shots 40 times out of 604 shots or 6.6% of the time. That’s every 18 shots on average or roughly every 4 days. I haven’t quite seen a plateau yet.

To me, the one true God Shot was my first shot. You can never repeat a first. I didn’t even know what I was getting into. I went to a cafe, and I ordered the cheapest drink on the menu. It happened to be espresso and marvelous. The sharpness of the first shot can never be repeated as now the palate knows.

My hope is that keeping a data sheet or shot journal could help you improve your shot; at least allow you to see what you like and don’t like. My categories are the things I’m looking for, but they don’t have to be the same for you. I collect extra data as well because you never know when you will need that data in the future.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Robert McKeon Aloe

I’m in love with my Wife, my Kids, Espresso, Data Science, tomatoes, cooking, engineering, talking, family, Paris, and Italy, not necessarily in that order.