Coffee Data Science

The Duality of Coffee: An Extract and a Filter

Undrinkable but entertaining

Published in

Towards Data Science

6 min readJan 22, 2021

Previously, I experimented with trying to turbo charge my espresso shot by feeding espresso back into the puck to extract more coffee. Instead, I found that used or spent coffee grounds filters coffee. The result was weaker espresso.

I wanted to better understand how something could both be used to extract and filter espresso.

So I devised an experiment based on staccato tamped shots. I’ve been pulling staccato tamped shots for almost a year, which simply means tamping half of the coffee grounds first, then the second half separately. I have also been using a paper filter in the middle, which means the top and bottom of the puck is ease to split after the shot is done.

Staccato Tamped: Generic Process Diagram

Experimental Design

With coffee grounds on top of spent grounds, we could measure just how much of the coffee is stuck in filtration.

Making Used Grounds

The key component to the experiment is having coffee that has nothing left to extract. I took a used puck, mixed it with a tooth pick, and ran it through the machine again.

Second round of extraction to pull out everything remaining.

Then I took that puck, mashed it on a tray, and put it in a toaster oven at low heat to dry it out.

I had to use two pucks to get 18g of used coffee. For both of those shots, I extracted close to 30% before drying. They then lost 40% of weight in drying, so a total weight loss of 50%. I’m not sure if it has any influence because each puck started 18g dry resulting in 9g of spent coffee.

Good on Top

Let’s take some shots! I did salami shots as well because I wanted to know the effects of EY throughout. This first shot hit some hard times, but I didn’t lose that much liquid.

First cup, poor transition to the second cup, second cup (third cup not shown)

Good on Bottom

The bottom went better, and you can see the starting liquid is darker.

First cup, First Cup still running, Second cup (third cup not shown)

All Unused Grounds

The full shot went quite well.

Experimental Data

The data shows some major issues with the grounds on top vs the bottom. The final EY might be lower on the Good on Top because of the spill, but it is certainly accurate for the first cup. I used a different VST basket for both, but I have previously shown no difference in performance between the two baskets (VST R for Ridged, RL for Ridgeless).

I plotted the data and made a best fit curve. Extraction is a natural log (ln) process. The Good on Bottom had a much higher extraction throughout indicating filtration is occuring.

We can combine the best fit for all three, and I estimated what the full shot should be in extraction if these two good halves were combined. It turns out that the Full Shot trended with the Good on Bottom shot.

The one challenge was that because I was only using 9g for each half, the shots were hard to control to determine when to use another cup. I even used a precision scale, but the change in weight from one second to the next was difficult to control. I would have preferred to have done this test on a Decent Espresso machine instead of a lever machine for that reason.

Drying Pucks

I also measured dried pucks. This only measures final extraction, but this way, we could see if anything was left in the used coffee or if any extract from the top stayed in the bottom.

I let all halves dry separately for a week. The Good on Bottom had used grounds on top which weighed dry at 9.05, so that’s a 0.56% difference. They should have retained no solubles, so I used that small percentage to correct the other data.

We can then look at some trends. The full shot has an even extraction by half than I thought it would.

I would have thought during a full shot would have had some filtration effects in the bottom half, but it doesn’t seem to have had such. I also thought if the bottom half was filtering coffee, for the case of Good on Top, the bottom half of the coffee grounds should have gained some weight after drying, but they did not.

While many of my experiments have seems strange and disconnected, I feel like they have been converging to a deeper understanding of the inner workings of an espresso 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. You can also follow me on Medium.