Gravimetric Brewing: Why Volumetrics Hit Their Limits with Fresh Coffee
Gravimetric Brewing: Why Volumetrics Hit Their Limits with Fresh Coffee
The common explanation goes: CO2 bubbles fool the Flowmeter. That is not quite how it works – the Flowmeter never sees the drink. The actual reason is more interesting, and it affects every volumetric machine, no matter how good it is.
A Flowmeter sits before the group head and counts pulses from a small impeller in the pressurised water. It measures clean water, never the Espresso.
The error therefore is not in the measurement, but in the quantity being controlled: volumetrics regulate the amount of water before the Puck, while what matters is the mass of the drink in the cup. In between sit retention, extraction and the gas content of the finished drink.
And that gas content is at its largest, and changes fastest, with fresh coffee. Crema makes up at least 10 percent of the Espresso volume and has a density of only 0.40 to 0.60 g/ml.
How a Flowmeter actually measures
Technically, a Flowmeter is a turbine or impeller counter. A small disc with vanes spins in the water flowing past it, and one of the vanes carries a magnet. On every rotation the magnet passes a sensor and generates an electrical pulse. The machine counts pulses and stops at the target value. Typical resolution: around 1300 pulses per litre.
La Marzocco also notes in its own documentation that Flowmeters depend on steady flow rates to be accurate. With very fine grind settings, water could historically even bypass the impeller, so that the rotations were no longer counted correctly.
That is physically impossible – the Flowmeter sits upstream of the Puck and sees nothing but clean pressurised water. The second version does not hold either: that CO2 makes the drink heavier. Let us do the maths: 18 grams of dose times around 6 milligrams of CO2 per gram gives about 108 milligrams. On 36 grams of drink that is less than 0.3 percent. The CO2 mass is irrelevant. What matters is the CO2 volume.
How much CO2 is really in coffee
| Roast level | Residual CO₂ in whole beans |
|---|---|
| light | 6.29–6.70 mg/g |
| medium | 11.04–11.51 mg/g |
| dark | 15.36–15.62 mg/g |
These values come from Wang and Lim (2014). The second part of their measurement matters just as much: whole beans need over 800 hours to release around 90 percent of their CO2. Ground coffee, depending on grind size, reaches a plateau after about 50 hours. And fine grounds lose 45 to 59 percent of the trapped CO2 during grinding alone.
Peer reviewIlly and Navarini (2011) measured 4.0 to 8.6 mg/g immediately after grinding, 5.7 mg/g on average, which corresponds to around 2.9 millilitres of gas per gram. It is exactly this gas that forms the Crema as it leaves the machine: in the Puck, at around 9 bar and 100 degrees, the CO2 is dissolved; on exit at 1 bar and 60 degrees the solution is roughly twice supersaturated and bubbles out.
Volume is not mass
This is where the real point sits. For the Crema, Illy and Navarini give a foam density of 0.40 to 0.60 grams per millilitre and put its share at at least 10 percent of the Espresso volume. A layer of Crema therefore weighs about half as much as the same volume of liquid.
A schematic illustration of the principle, not a measurement. What is documented is the direction: Wang, Lim, Tan and Fu (2019) examined seven parameters and found that Crema volume depends above all on variety and freshness – fresher coffee consistently produces more Crema. Particle size, brew ratio and temperature barely mattered.
So the clean statement is not «volumetrics measure wrongly», but: volumetrics regulate the wrong quantity. They regulate the volume of water before the Puck. Between that water and the cup sit Puck retention, extraction yield and the gas content of the drink – and none of those quantities is regulated. How freshness and roast level play into this we have described under Why we do not give a roast date in days.
What has actually been measured about variance
Here we have to be honest: the evidence is thinner than the marketing claims suggest. We found no figure – neither peer-reviewed nor from a clean manufacturer test – showing that a gravimetric cut-off reduces Shot-to-Shot variance by any specific amount.
Manufacturer labWhat does exist is a test series by La Marzocco itself with four conditions, target 20 grams in, 40 grams out, 18.5 percent yield:
| Method | Mean yield | Standard deviation |
|---|---|---|
| time only (28 s) | missed the target in every category | |
| sight + time | 18.30 % | 0.52 % |
| volumetrics, dose only roughly set | partly above 19.00 % | |
| volumetrics + dose weighed to 0.1 g | 18.54 % | 0.30 % |
Read strictly, this test does not compare volumetric against gravimetric drink cut-off at all. It shows something else, and we find that almost more important: dosing accuracy at the grinder is the biggest lever. Notable, too, that it is a maker of volumetric machines reaching that conclusion.
Our reading of it: if you can only change one thing, weigh the dose. If you can change two, weigh dose and drink. That is not science, it is a prioritisation based on the best data available. What a precise scale changes in practice we have described under Precision in the cup, and which scale suits whom under Acaia vs. Normcore.
The same problem at the grinder
At the grinder the same logic repeats: timer dosing regulates grind time, not mass. What sits in between is bean density, hopper level, retention and residual moisture.
ManufacturerThe most concrete figure on this comes from Lauro Fioretti of Simonelli Group: between a full and an almost empty hopper, at identical settings, up to 1.5 grams of difference can appear. On top of that, with timer dosing, grind setting and dose are mechanically coupled – go finer and you change the dose automatically.
Modern Grind by Weight grinders quote dosing accuracies of around 0.067 grams to ±0.1 grams. These are manufacturer figures, not independently verified and not documented methodologically – we quote them as an order of magnitude, not as a measurement result. You will find the overview of the models under HOME Grind by Weight grinders compared and the distinction from Single Dosing under Grind by Weight vs. Single Dosing.
One claim we deliberately do not make: that fresh coffee runs slower because of CO2 bubbles in the Puck. There is a good counterargument – at 5 bar in the Puck such bubbles would be only around 0.4 micrometres across, small compared with the pores, and on paper the water could dissolve more CO2 than is present in the Shot at all. We found no solid study measuring Shot time against roast age. An open question, and we are leaving it open.
Frequently asked questions
Does a Flowmeter measure CO₂ bubbles too?
No. The Flowmeter sits before the group head in the pressurised water circuit and measures only clean water, before it touches the coffee. It never sees the finished drink. The real point is a different one: from the same amount of water, different coffees produce different drink masses.
Why is gravimetric brewing more accurate than volumetric?
Because it regulates the quantity that actually matters. Volumetrics regulate the volume of water before the Puck. In between sit Puck retention, extraction yield and the gas content of the drink. A concrete figure for how much the variance drops, though, is neither peer-reviewed nor documented in a clean manufacturer test.
How much CO₂ does freshly roasted coffee contain?
Depending on roast level, around 6.3 to 6.7 mg/g for light, 11.0 to 11.5 mg/g for medium and 15.4 to 15.6 mg/g for dark roasts. Whole beans need over 800 hours to release about 90 percent of it. Fine grounds lose 45 to 59 percent during grinding alone.
Does fresh coffee make more Crema?
Yes, that is documented. Wang, Lim, Tan and Fu (2019) examined seven parameters and found that Crema volume depends above all on coffee variety and freshness. Because Crema, at 0.40 to 0.60 g/ml, is only about half as dense as the liquid phase, more Crema at the same volume means less mass.
Which gives more: a scale on the machine or on the grinder?
Going by the best available test series, the grinder. In a lab test by La Marzocco, dosing accuracy at the grinder was the single biggest lever – the standard deviation of the yield fell from 0.52 to 0.30 percent once the dose was weighed to 0.1 grams. That is a manufacturer measurement, not a peer-reviewed study.
Roastery tip
If you want to see the effect yourself: set the machine to volumetric, pick a freshly roasted coffee and weigh along anyway. With Wild Peach, which is very lively in its first days, you will see the deviation fastest. With Very Nutty, which is calmer, it turns out smaller. Two coffees, two results, the same machine – that is the whole point. Our recommendation, one of many.
