Put a batch of green coffee on the scale, roast it, weigh it again, and the number will be smaller. Always. Every roast in every machine on every continent ends lighter than it started, and the size of that gap is one of the most useful numbers a roaster can record. It is called roast weight loss — also shrinkage, roast loss, or the inverse of roast yield — and it is simply the percentage of mass that left the batch between charge and discharge.
What makes it valuable is that it takes almost no effort to collect and it is very hard to fudge. A scale and a notebook are the entire toolkit. Yet the number quietly encodes green moisture, how hard the roast was driven, how far development ran, and whether today's batch matches yesterday's. This guide covers what the loss is made of, the percentage bands to expect at each roast level, how to calculate it cleanly, and why a half-point drift is worth investigating.
What roast weight loss actually measures
Roast weight loss is a mass balance, not a flavor score. It tells you how much matter left the bean as vapor and gas, and nothing directly about how good the coffee tastes. Two roasts can land on the same loss figure and cup completely differently if one was rushed and the other stalled. Treat it as a consistency instrument rather than a quality verdict.
It is also distinct from the green specification you started with. The moisture reading taken on arrival belongs to the green coffee and is covered separately in our guide to green coffee moisture content. Roast weight loss is what happens to that coffee once heat is applied — it consumes the moisture figure as an input and adds a second, heat-driven component on top of it.
The two components: water first, then dry matter
Loss is not one process. It is two, running partly in sequence and partly in overlap.
Component one — water
Green coffee is generally stored somewhere in the region of 10–12% moisture, and nearly all of that water is driven off during roasting — finished roasted coffee is generally cited as retaining only about 1–2% moisture. Research reported in the specialty trade press found that beans shed more than 70% of their initial moisture within roughly the first six minutes of a conventional drum roast, with the remainder trailing off as the bean heats through. Because the water is baked out almost completely regardless of how dark you go, this component sets a floor: a coffee that arrived wetter will lose more mass than a drier lot roasted identically. That relationship is close to one-for-one, which is why roasters who track moisture alongside water activity in green coffee can anticipate their loss figures before the first batch is even charged.
Component two — dry matter
Once the bean is dry, further heat starts consuming the seed itself. Sugars caramelize, chlorogenic acids break down, cellulose pyrolyzes, and the bean vents carbon dioxide plus a stream of volatile compounds and newly formed water of reaction. This is real substance leaving the coffee, not just packaging water, and it accelerates the longer you stay past first crack. Dry-matter loss is the component you control with the roast profile; water loss is the component the green coffee hands you.
A workable mental model: total loss ≈ green moisture + roast degree. The first term is fixed by the lot. The second is fixed by your decisions in the drum.
Typical percentage bands by roast level
Published figures vary between sources because machines, batch sizes, airflow and green moisture all differ, so treat these as orientation rather than targets. Most specialty roasters working from light to medium report landing somewhere between roughly 11% and 16%, while the full range across all roast styles is generally cited at about 11% to 24%.
| Roast level | Typical weight loss | What is driving it |
|---|---|---|
| Very light, dropped at or just after first crack | Roughly 11–13% | Almost entirely water; minimal dry-matter loss |
| Light-medium filter | Roughly 13–15% | Water plus a modest slice of dry matter |
| Medium to medium-dark, espresso-oriented | Roughly 15–18% | Meaningful pyrolysis and carbon dioxide generation |
| Dark, into or past second crack | Roughly 18–20%+, sometimes into the low 20s | Heavy dry-matter breakdown; oils migrate to the surface |
One consultancy heuristic worth knowing: some importers and roasting consultants report that lots landing in an 11–13% loss window tend to show the cleanest flavor development, with lower figures skewing grainy and underdeveloped and higher figures skewing bitter and flat. That heuristic is specific to their coffees and roasting style, but it illustrates how loss gets used as a proxy in practice.
How roasters calculate shrinkage
The arithmetic is trivial and worth writing on the wall above the scale:
Weight loss % = (green weight − roasted weight) ÷ green weight × 100So a batch charged at 10.00 kg green and discharged at 8.50 kg has lost 1.50 kg, which is 15.0%. Roast yield is the mirror image of the same measurement — 85.0% in that example — and some operations prefer to log yield because it maps more intuitively onto production planning. Either convention works as long as the whole team uses one.
Getting an honest number
Sloppy measurement destroys the signal, and the errors are usually procedural rather than mathematical:
- Tare the same containers every time. A hopper weighed with its lid on one day and off the next introduces more variation than the roast itself.
- Weigh at a consistent point after drop. Coffee keeps venting carbon dioxide for days, and it also reabsorbs a little atmospheric moisture. Weighing straight off the cooling tray versus twelve hours later gives different answers. Pick one and stick to it.
- Account for chaff and QC pulls. Silverskin blown into the cyclone is genuine mass leaving the batch, and if you scoop samples out for cupping or color reading before weighing, you inflate the apparent loss.
- Log green moisture per lot. Without it you cannot separate a green change from a profile change.
Why it is a quality-control signal
Here is the payoff. Run the same coffee, on the same machine, with the same profile, and the weight loss should repeat closely. When it does not, something moved. A widely used shop-floor tolerance is half a percentage point: a batch sitting 0.5% or more away from your established baseline for that coffee deserves a look before it goes into a bag. In practice, batches that breach that band very often also read visibly off on whole-bean or ground color.
Interpreting a drift is a process of elimination:
- Loss suddenly higher? Wetter green than the last shipment, a hotter charge, more aggressive airflow, or a later drop. Cross-check against drop temperature before blaming the coffee.
- Loss suddenly lower? Drier green from a lot that has been in storage a while, an underpowered burner, an overloaded batch, or a drop pulled early.
- Loss creeping across a shift? Classic thermal-soak drift as the machine heats up through consecutive batches — the fourth roast of the morning behaves differently from the first.
Weight loss becomes far more powerful when triangulated with the other two cheap measurements. Read alongside a color reading, it separates "roasted darker" from "roasted wetter" — a batch that lost extra mass without darkening on the Agtron scale almost certainly started wetter rather than going further. Read alongside your development time ratio, it distinguishes a genuinely extended finish from a batch that simply drove harder through the drying phase. Three numbers, three different failure modes, and none of them requires equipment beyond what most roasteries already own.
The paradox: volume rises while mass falls
The counterintuitive part of shrinkage is that the beans do not shrink. They swell. Over a typical roast the seed expands markedly — reported figures range from roughly half again to about double the original volume, depending on how far the roast is taken — driven by steam and carbon dioxide generated inside a cellulose matrix that softens under heat and inflates like a slow, rigid balloon. Mass goes down; volume goes up. Both are true at once, and the word "shrinkage" refers only to the mass side of the ledger.
The consequence is a steep fall in density, and it is worth separating the two ways density gets reported. At the level of the individual bean, roasted coffee is commonly reported around 0.7–0.8 g/cm³ for lighter roasts and nearer 0.5–0.6 g/cm³ for darker ones, down from a green figure generally cited well above 1 g/cm³. Bulk density — loose beans in a container, which is effectively what a scoop measures — runs lower still, with published figures often falling somewhere in the region of 0.33 g/ml for very dark roasts up to around 0.45 g/ml for very light ones. Both measures move the same direction as the roast progresses. This is exactly why the density sorting applied to green lots — see density grading — describes a completely different population than the airy, porous, fragile bean that comes off the cooling tray. It is also why a scoop is a poor unit for roasted coffee: the same scoop of dark roast holds noticeably less coffee by mass than the same scoop of light roast.
What it means for dosing by weight
For anyone brewing, the practical translation is simple. Dose by mass, never by volume, and recognize that a fixed dose of dark roast contains more beans, more surface area, and a more porous structure than the same mass of light roast. That is a large part of why dark roasts extract faster and why grind settings rarely transfer cleanly between roast levels.
On the production side, weight loss is the bridge between what a roastery takes in and what it can pack out, since green arrives by mass and roasted coffee leaves by mass. Every point of loss is real product that evaporated. That is not an argument for underroasting — a stalled, underdeveloped batch is worth nothing regardless of how much it weighs — but it does explain why the figure is tracked so closely wherever repeatability matters. If you are still assembling the mental picture of what happens between charge and drop, our overview of how coffee roasting works lays out the phases that produce this mass loss in the first place.
Start logging green weight and roasted weight for every single batch. Within a few weeks you will have a baseline per coffee, and from then on the scale becomes an early-warning system that catches inconsistency well before the cupping table does.