The Diminishing Returns of Tamping for Espresso

A short experiment to better know your pressure

Published in

Towards Data Science

7 min readAug 14, 2020

This experiment started from a Facebook discussion in an espresso group about tamping vs tapping. The thinking was tamping compresses grounds at the top and the bottom more than the middle, and this is why one should use a heavy tamp. So I devised an experiment to better understand tamping.

Short story: tamping affects the top layers more than the bottom layers with respect to compression. As tamp pressure increases, there is a diminishing returns on puck compression that is non-linear.

Previous work has shown that tamping pressures at 5 lbs to 20 lbs doesn’t show a change in extraction. My own experience has been to greatly drop the tamp pressure, but I didn’t have an understanding on how pucks were compressed. I was also not able to find much information online.

So, I setup this experiment with a straight, clear tube and a scale to see how the grounds would compress. I used Kompresso for the tube because it has a clear, straight, and long with a plunger to fit. I put it on a scale that could go up to 5 kg, and I recorded a video as I pressed down.

I did the collection twice because the first round had some trouble in being able to label the markings, but the second round went a lot better.

The two questions to resolve:

How does tamping affect the top, middle, and bottom of the puck?
When does a harder tamp provide diminishing returns?

Data Collection

I put a small piece of paper after every 3g of coffee in the tube, but I aimed to keep the paper small. I didn’t want the piece of paper to affect how the coffee compressed. Putting the initial grinds in the tube did not cause compression as can be seen above.

I took sample images at desired pressures, and then I lined them up with the starting image. I then used red lines for the starting positions and green lines for the ending positions.

I wrote a simple script in Matlab to find the location of red and green lines in the images.

Calibration and Labeling

I also wanted to remove camera bias so I used a ruler to make a calibration image. I then marked it up with red lines and used the same script to find the locations of the red lines. I then fit a line to these markings and converted the line locations for each image to a physical distance.

For each image, I was able to clearly get four lines for the slices of the Top, Middle, and Bottom. There was coffee above the Top and below the Bottom, but I can’t see the top, and I’m suspicious about using the bottom. I ended up using the three slices and the four lines (used to derive the slices) in analysis.

The plunger is also smaller than a 58mm (58.5mm to be exact) filter basket. To accommodate, I scaled the pounds per square inch from the 0.93 square inches of the Kompresso tube to the 4.16 square inches of a 58 mm basket.

Tamp Pressure Analysis

I plot here the slice height and then the height of the markers. I plotted the height of the markers in log to show they trend logarithmically. This means that to gain the same change in height change or compression, the force required grows exponentially.

There is some noise in the data, but the trend is pretty clear. When looking at a line plot, we can see the trend is that the Bottom doesn’t compress like the Top and Middle. We can look at the percentage of compression with respect to the maximum compression achieved at 59 lbs.

The Top compresses mostly in the first 6 lbs, and the middle is the slowest to compress. This makes sense because when you compress the grounds, there will be an equal and opposite force from the counter on the bottom grounds. In absolute values, the bottom is compressed less. This could be due to the paper marker moving when the grounds were put in. This experiment would work better if one set of grounds was colored and one wasn’t. My aim here was to get a quick experiment to see what it could tell me about tamping.

Comparing Slices

We can use the slice compression percentage and compare different slices. I did this by making a ratio of compression as seen below. The Top/Bottom ratio is the highest initially, which seems a bit obvious.

Looking closer, the Top/Mid and Top/Bottom seem to even out while the Mid/Bottom stays below 1 which means that the Bottom compressed more than the middle.

Focusing on the Line Markers

If we look at how much the line markers move as a percentage of their total movement or compression, we can see a little more on the middle. The Mid-Top and Mid-Bottom trend the same way. When comparing one layer to the next, they have close to a 1:1 ratio. However, when looking at the different lines vs the Bottom line, all the layers compress faster than the bottom up until 15 lbs. Then they are all fairly close (within 20%) to their maximum compression positions.

Diminishing Returns

If we look at the overall height of the puck, the idea of diminishing returns is very clear. 80% of compression happened at 3 lbs of pressure. To get to 90% compression took 20 lbs, and then to get to 100% compressed took more than 50 lbs. 50% of all compression occurs in the first 1 lbs of pressure.

I couldn’t get it to compress more than this point, and I wasn’t able to measure pressure past 13 lbs with this scale or the equivalent of more than a 60 lbs tamp for a 58mm basket. However, I doubt doing a 60 lbs tamp makes much of a difference. Previous experiments on tamp pressure showed there was no difference in extraction when comparing 10 lbs of tamp pressure t0 40 lbs. They didn’t go lower than 10 lbs though, so it is unclear.

In my own experiments, I’ve been reducing my tamp pressure to 1 lbs or less.

A better experiment can be designed as a result of this one:

Use two different colors of coffee grounds
Layer at 2g increments
Look at the compression.

I am satisfied in now having a better understanding of our two questions:

1. How does tamping affect the top, middle, and bottom of the puck?

Not evenly even at very high pressures.

2. When does a harder tamp provide diminishing returns?

Between 2 lbs and 3 lbs of pressure, there was a 20% increase in compression compared to the 10% between 3lbs and 20lbs or the 10% between 20lbs and 59 lbs. This means that most tamps heavier than 3 lbs should be similar, and that a massively heavy tamp may not be worth whatever effect it is having on the shot.

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.