I roasted for about a year in a state I'd describe as cheerfully random. Some batches were lovely. Some were flat. I couldn't reliably tell you why, because my whole record-keeping system was a stopwatch and a hunch about bean color under kitchen light. What fixed it wasn't a better roaster or a thermocouple. It was a fifteen-dollar kitchen scale and one derived number: how much mass the beans lost between green and roasted. That single percentage turned my roasting from a series of one-off attempts into something I could actually repeat on purpose.
Why color and the timer both lie
The two things a beginning roaster naturally leans on are the least dependable references in the kitchen. Color is deceptive because it depends on the origin, the process, the light you're standing under, and whether you're looking at whole beans or ground. Two coffees pulled at the same internal development can look a full shade apart.
Time is worse, because it measures the wrong thing entirely. A stopwatch records how long your appliance ran, not how far the beans got. Ambient temperature, batch size, line voltage, and how warm the machine was from the previous batch all shift what nine minutes accomplishes. I've had two popper roasts of the same coffee, same duration, land in visibly different places because the second one started with a hot chamber. The clock said they were identical. My mouth disagreed.
Weight loss sidesteps both problems. It measures what actually happened inside the bean, after the fact, in a unit you can't argue with.
The arithmetic, done on a kitchen scale
The formula is grade-school subtraction and division:
loss % = (green weight − roasted weight) ÷ green weight × 100
Weigh the green beans before they go in. Weigh the roasted beans once they're fully cooled and you've shaken the chaff off. Then record it. A 100-gram green charge that comes out at 86 grams lost 14 percent. No probe, no software, no curve.
A few mechanics matter more than they sound like they do. Use a scale that reads to a tenth of a gram; on a 100-gram home batch, that's a tenth of a percentage point, which is plenty of resolution. Weigh a round green charge every time, because working from 100 grams makes the mental math instant. And always weigh cooled, never hot — beans keep shedding moisture and gas while they cool, so a warm weight understates the loss by an inconsistent amount.
Chaff is the one judgment call. The silverskin that flakes off during the roast is real mass that leaves the bean, and in an air popper most of it blows out of the chamber on its own. My rule is to weigh beans only, chaff discarded, and to be consistent about it batch to batch. Whether you count chaff matters far less than whether you count it the same way every time.
What the percentages actually correspond to
The range you'll live in is narrower than you'd guess. Royal Coffee's discussion of roast yield and loss frames the outer edges usefully: a very light roast dropped right at the onset of first crack loses only around ten percent of its green weight, while a roast taken well past second crack can shed as much as a quarter of it (royalcoffee.com). Their shorthand for what drives the number is that roast loss is green moisture plus roast degree, which is a tidy way to say two things are leaving: water, and the products of the bean cooking.
That gives you a map. Everything home roasters want sits between those poles, and most of my keepers land in the low-to-mid teens. Below about eleven percent I'm usually tasting something underdone — grassy, sharp, thin in the middle. Past the high teens the origin character flattens into generic roast flavor, a legitimate choice for a dark, chocolatey cup but not what I want from a good single origin.
Where the vanished mass goes
It helps to know you're not measuring an abstraction. Green coffee arrives holding meaningful moisture — commonly cited in the neighborhood of ten to twelve percent by weight — and driving that water off is the bulk of what disappears in the roaster. The rest is chemistry: as the bean browns and develops, it gives up carbon dioxide and a crowd of volatile compounds, and it physically expands while doing it.
Which is why loss tracks development rather than just drying. If it were only water leaving, the number would top out near the green moisture content and stop. It doesn't, and the amount past that ceiling is a rough proxy for how far the roast went. It also explains a fact that surprises new roasters: roasted beans are lighter but bigger, which is why a scoop is a useless measure and a scale isn't.
Turning it into a roast log
My log is a cheap notebook with one line per batch, and it holds exactly six things:
- The coffee — origin and lot, whatever the bag says.
- Green weight in, always the same round number.
- Roasted weight out, cooled.
- Loss percentage, calculated on the spot before I forget.
- Total roast time, plus the minute first crack started if I heard it.
- One sentence about the cup, written a day or two later.
Six fields, thirty seconds of writing, and after a dozen batches you own something no guide can give you: a record of what your specific setup does. Mine told me my popper drifts about a percentage point darker on the second consecutive batch, a real bias I now correct by pulling batch two early.
The habits that corrupt the number
A few things will quietly make your percentages meaningless. Changing green batch size between roasts is the big one — a 70-gram charge and a 120-gram charge in the same appliance behave like different roasters, and comparing their losses tells you nothing. Weighing hot beans is the second. Storing green coffee somewhere humid is the sneaky third, because green moisture is half the equation, and beans that have sat in a damp basement start from a different place than the same lot did in January.
The last one is discipline: calculating the number and not writing it down. I did that for weeks, told myself I'd remember, and remembered nothing. The value here is entirely comparative, and a figure you can't compare against anything is just trivia.
Two coffees, same percentage, different cups
Be clear about the limits. Weight loss tells you roughly how far a roast went, not how well it got there. Two batches can land on identical losses and taste nothing alike if one raced through development and the other crawled — the drop is the same, the path was not. It also won't compare cleanly across coffees, since a dense, low-moisture washed lot and a soft natural won't sit at the same percentage at the same perceived roast level.
So it's a coordinate within one coffee on one machine, not a universal score. That's still far more than color and a stopwatch were giving me, and it's the foundation everything else gets built on top of.
What I do every Saturday now
The ritual costs maybe a minute of extra effort. I weigh out 100 grams of green, note the coffee, roast, cool the beans completely on a mesh screen, shake the chaff loose, and weigh again. I do the division in my head, write the percentage next to the weights, and leave a blank line for the tasting note I'll add midweek. If a batch was excellent, I know what to aim for next time. If it was flat, I know which direction to move and roughly how far.
If you're roasting without this, start on your next batch. Two weighings and one division, logged in a notebook you keep next to the roaster. It won't make a single batch better on its own — but it's the difference between hoping the good one happens again and knowing how to make it.