Coffee Data Science

A Response to Lance Hedrick’s Video on the Ultimate Guide to Refractometry

A different view of the same data

Robert McKeon Aloe
Nerd For Tech


Lance Hedrick came out with a video explaining refractometry and then comparing four refractometers to each other. He compared them using filter and espresso brews, and he made the data available online. He looked at 4 refractometers: VST, Atago, DiFluid, and Yellow Boi. I was curious about data, and I found a conclusion contrary to his supported by his data.

He took multiple brews and sampled using each refractometer. He found the Yellow Boi to be inaccurate assuming VST was the most accurate. He also found the DiFluid to not be accurate. Additionally, he had a concern was results was changing if with too much noise if he took multiple samples.

In using the DiFluid, he used two or three drops of coffee, and he found the sample would increase in TDS if it repeated the measurement. He assumed water evaporation was negligible, but he did not go deeper into understanding what was occurring.

The Data

I previously reviewed the DiFluid using data, and I found for espresso, it had a similar distribution as the Atago if the samples were cooled. I also provided some data to help show that the change in TDS was due to a calibration variable related to temperature.

On the other hand, Lance found the DiFluid to not match well with the VST. The main way he presented his results was with a box plot. I love box plots, but in my experience, with small data sets, plotting the data is better as simple as possible especially for paired data.

So I put the same data into scatter plots. For filtered coffee, the Atago and VST nearly match, which matches previous research by Socratic Coffee. However, DiFluid produces higher results and Yellow Boi produces lower results. The DiFluid appears to have a slight trend (R² of 0.81), but there are too few samples to be sure. My plots only have 9 data points even though Lance took 10, but only 9 were available in the data he shared.

All images by author, the black line means the sample has the same reading as the VST.

Espresso shows a different picture where DiFluid trends very strongly with Atago and VST. Yellow Boi has an offset.


To understand these distributions, we should run a statistical t-test. We can run a two tailed, paired t-test to evaluate the null hypothesis that the two distributions have no statistical difference. A p-value less than 0.05 means that two distributions are not the same. This test is really strong and helpful in the case where the data is paired like in this case.

We can look at filter and espresso separately and together. Filter by itself shows statistically different results between DiFluid and VST/Atago. This is not true for espresso where the distributions for VST, Atago, and DiFluid are not statistically different. Yellow Boi is statistically different from all.

Green indicates a p-value < 0.05 which indicates the difference in the distributions is statistically different.

The same holds true for the all results where VST, Atago, and DiFluid don’t have distributions with statistically significant differences from each other.

Lance’s original suggestion was to get an Atago or a VST (if you have more disposable income). However, I think when using DiFluid for espresso works fine, at least based on his data. I suspect more data is needed for filter coffee. I would have preferred a larger data set to better understand the differences, and for filter, I’m sure someone is up to the task.

If you get a DiFluid, I suggest using more than a few drops. I fill the resevoir for consistency, and that helps with any evaporation. I would also suggest:

  1. Cooling the sample.
  2. If possible, cover the cooled sample, but don’t cover a hot sample. Water could condensate on a cool surface, thus raising the TDS results.

If you like, follow me on Twitter, YouTube, and Instagram 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 and Subscribe.

Further readings of mine:

My Future Book

My Links

Collection of Espresso Articles

A Collection of Work and School Stories



Robert McKeon Aloe
Nerd For Tech

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.