It’s no secret that I love a good density reading. In terms of ease of measurement and how much it can tell you about your green coffee, the metric is unmatched. 

I’m also a giant nerd for amateur data sleuthing, and looking back over many years of Crown Jewel data recently led me to wonder a little about how precise and repeatable our practices are. 

I decided to compare four measurement techniques, using five coffees at varying densities, and see if I could find anything meaningful. 

It turns out there are some interesting takeaways: 

    1. I was able to confirm a high degree of repeatability and a low percentage of variation in our current methods. 
    2. I found some interesting, albeit minor differences between density measured in a cylinder versus in a cup, as well as between my manual techniques and the digital readings on the Sinar BeanPro 
    3. I was unable to achieve sufficient precision in multiple displacement-method density readings 
    4. I feel very confident in the precision of our current methodology and continue to believe that it can be easily repeated by any user with a bare minimum technology and investment. 

Current Practices 

We currently use two methods regularly for our Crown Jewels: manual free settled measurement in a 100mL graduated cylinder, and digital free settled measurement in a Sinar BeanPro moisture meter. (Free settling, or bulk density, is the preferred measurement method for coffee).  

The graduated cylinder has proved reliable and consistent over time. I tend to trust those measurements for two main reasons: first, our methodology is consistent, and second, we take multiple readings of each coffee and average them to mitigate operator error. 

The cylinder has markings up to 100mL, but we’ve found it much easier and more consistent and reliable to fill the cylinder to the top with coffee and use a straight edge to level off the top. Without millimeter markings, we were able to determine the total cylinder volume by filling it with water and weighing the contents. For practical purposes, water mass in grams is equal to the volume in milliliters, so we know that the volume to the top of our tool is exactly 136.5 mL. 

The Sinar BeanPro also measures density in this way. Internally a balance takes the mass of the sample and then divides by the volume of the cup that accompanies the device, whose volume is approximately 270-275mL. 

The sample cup is both larger and much wider than a graduated cylinder, so it’s not a huge surprise to find that the figures differ from time to time. The most likely common cause of this discrepancy is likely bean size and shape. Oblong or extra-large or small beans, for example, will naturally fall into the differently shaped spaces irregularly. Because of how narrow the graduated cylinder is, it’s likely that these irregularities will be magnified using this method. 

Both methods are also susceptible to operator error. In some ways, the larger sample cup is possibly less so due to the larger sample size. However, the wide opening also allows for greater error potential when filling and leveling. In either case, consistency and repetition are widely acknowledged keys to success. 

Manual and Digital Free Settled Measurements 

Let’s look at a real-world example of this:  

a graph comparing the precision of various density measurements

This chart represents five different coffees across a range of free settled density readings. Measurements were taken three times per device: Graduated Cylinder, Sinar Cup (manually weighed), and the Sinar Digital Readout. 

I measured about 275mL of space in the Sinar cup, but it appears the Sinar digital calculation is based on a figure that ranges from 267-270mL depending on… what I’m not exactly sure. That difference in measurement mostly accounts for the higher figures represented on the chart. The cup, with its larger capacity and wider shape, also skews a little higher in final reading than the 136.5 mL graduated cylinder. 

As far as consistency, all three methods are pretty close. The Sinar digital readout had the largest variation on the Colombian coffee, with a 2.5% difference between the first and third reading, followed by the graduated cylinder with a 2.3% difference on two readings of the Sumatra. The manual cup measurement seemed most consistent when comparing maximum differentials. However, when averaging all variations, the graduated cylinder ended up as most consistent.  

In all cases, the weight differences that caused variations were minor and by taking averages of multiple readings, we can easily account for outliers. 

As far as interpreting the differences between measurements, I think the most important takeaway is that if your methodology is consistent (as ours has been for years) then by gathering sufficient data with enough samples, you can establish reliable benchmarks. 

In other words, for the graduated cylinder technique, we can be certain that a measurement of 700 g/L is very high, 680 g/L is about average, and 660 g/L is fairly low. However, if I were to use Sinar’s digital readout, a measurement of 730 g/L would be very high, 700 g/L might be a little closer to average, and 670 g/L would be on the low end. It’s all about the consistency of your technique and your device. 

Displacement Difficulties 

Volumetric displacement is not a recommended method for measuring a coffee’s density, in part because it requires submerging the beans in water, effectively ruining them. It’s also not fully representative of the way we typically pack or store coffee, so its usefulness is limited. The surface tension of water also makes calculating this measurement somewhat problematic despite the theoretical ease of technique. 

That said, I wanted to take another pass at measuring using displacement, however, to try and confirm the free settled method’s results. 

I ran into immediate problems. First, the 100mL graduated cylinder (despite 1mL markings) was ultimately too small of a vessel. 1/10th of a gram increment in added coffee was insufficiently precise to correctly calculate, and lack of more than 1mL accuracy of reading meant that my results were all over the map.  

I decided to scale up to a 250mL beaker with 50mL markings, which is incredibly imprecise, but instead of adding a fixed volume of coffee, I decided that I could add 150mL of water and then add coffee until it reached the 250mL marking. 

This also proved somewhat challenging, as the marking on the beaker at 150mL ended up holding about 140g of water. Ugh. 

As a result of these frustrations, I think the biggest takeaway from my testing was that I needed far more precise testing equipment to make meaningful observations using displacement. As a result, I definitely do not recommend this method as an everyday technique for a quality lab. 

I ended up confirming with some degree of repeatability a few measurements with a marginal degree of confidence: 

    1. I confirmed that all the green coffee I used was denser than water. 
      • The least dense Yemen did indeed sink and returned a measurement of 1020 g/L (It took 102g to move the water level 100mL) 
    2. I confirmed that, yes, some coffees are denser than others. 
      • The high-density Kenya and Colombia returned measurements of 1070 g/L (107g to move the water level 100mL) 
    3. I found that generally my ability to distinguish meaningful results was impaired by both my imprecise equipment, but also by the method itself. 
      • Meaningful distinctions between density seem to be more apparent using consistent free settling techniques. 

green coffee beans


Conclusively, it’s simpler, more representative, and likely more accurate with consumer-grade lab equipment to measure a coffee’s density by free settling. 

I’d highly recommend using a version of this in your QC lab, as the metric will help you predict the coffee’s behavior both on your shelf and in your roaster. 

For further reading, see my article on Green Coffee Density, which appeared as a section in our four-part series on Green Coffee Analysis. A narrated PowerPoint presentation of this article is available on our resources page as well.