Coffee roast defects

Five faults one laboratory made on purpose

Sourced

A washed Kenyan arabica, roasted a kilogram at a time on one drum roaster. The reference roast began with the air at 210 °C and ran 11 minutes 25 seconds, with 2 minutes 40 seconds after first crack. Each fault changed that recipe as follows.

These times and temperatures are how this one study chose to produce each fault, on its coffee and its roaster. They are not thresholds, and the authors warn that the marker compounds may change with the green coffee.

The same word, two recipes

Accounts differ

Set that table beside another study and the vocabulary comes apart. A second research group, roasting on a larger drum roaster, also made a baked roast. Its version moved quickly to the colour change and then stretched the phase between colour change and first crack. The first group's baked roast did nearly the opposite, stretching the time after first crack. Both called the result baked, and the second group's paper notes that this shape is thought to reduce sweetness and create flat flavours.

Underdeveloped is looser still. In the first study it means a roast started cool and run slowly for twenty minutes. In everyday use among home roasters it usually means a roast stopped too soon after first crack, which in that study's terms is the light fault. A reader who is told a coffee is underdeveloped has been told that somebody disliked it in a particular way, and little about what was done to it.

None of this makes the words useless. It means a defect name is a description of a result, and the cause has to be worked out for each machine.

Can anyone taste the difference?

Sourced

The first group followed its chemistry with a tasting study, read here in the authors' manuscript. A trained panel separated the light roast clearly from the dark and scorched ones: dark and scorched were rated higher for intensity, bitterness, astringency and burnt flavours, and the light roast lower for all of them. The reference, baked and underdeveloped roasts sat close together, and the panel's differences between those three were not statistically significant.

Eighty-three consumers then scored how much they liked each coffee on a nine-point scale. The reference roast scored highest at 6.0 and the light roast lowest at 4.2, and the authors note the range was not large. The reference was liked significantly more than every fault except the baked roast, which scored 5.7. Consumers did describe the baked roast as more bitter and less sweet than the reference. The authors call their results inconclusive on how the baked and underdeveloped roasts differ from a normal one.

That is evidence that roasting far too light, too dark or scorched is perceptible and less liked. For the two faults roasters talk about most, baked and underdeveloped, the one study that tested them found them hard to tell from the reference. It is one coffee and one panel, and the faults were made by one laboratory's recipes.

Scorching, tipping and uneven batches

Widely documented

Scorching, meaning burn marks on the surface of the bean, has the firmest footing of the craft terms because it follows from how a roaster works. A doctoral thesis gives the purpose of every roaster's mixing mechanism as roasting evenly and preventing the scorching of beans, and its review reports that in very fast roasting the inside of a bean is less roasted than the outside.

Tipping, meaning burn marks at the ends of the bean, is a roasters' term for which CoffeeHQ read no source at all. It is mentioned here so that the word can be recognised, not explained.

For a home roaster the usual practical cause of an uneven batch is the load. One home drum roaster's manual warns that using less than its recommended amount of coffee makes over-roasting more likely, and each of the three makers' manuals read states the load its machine is built for. A batch the machine was not designed around does not move through the heat as intended.

Four kinds of thing a defect name can be

CoffeeHQ explanation

The same vocabulary is used for four different kinds of thing. Knowing which one a word is being used for settles most arguments about roast defects before they start.

A mark you can see on the bean

Scorching and tipping, in the sense of dark patches on the flat face or burnt ends, and an uneven batch with dark and pale beans together. These are physical defects: two people looking at the same beans will agree whether they are there.

A word for a taste

Baked, smoky, grassy, ashy. These are descriptors. In the one study that asked, a trained panel separated a light roast clearly from dark and scorched ones and could not reliably separate baked or underdeveloped roasts from the reference. A descriptor records what somebody tasted; it does not let a taster work back to the cause.

A diagnosis read from a curve

Baked because the rate of rise fell, underdeveloped because the time after first crack was short. This is the trade's reasoning from a shape to a fault. One paper notes that, to its authors' knowledge, no published data supports the belief that a negative rate of rise degrades quality.

A mechanism somebody demonstrated

The rarest kind. One laboratory produced five faults on purpose and found marker compounds for them, which establishes that those recipes give those chemical signatures. Even there the names are the laboratory's own: its baked roast stretched the time after first crack, and another laboratory's baked roast did nearly the opposite.

If a roast tastes wrong, what can and cannot be done

Tendency, not a guarantee

The flavour words here are common usage among roasters, not findings of the studies above. The point of each row is what to change.

Harsh, ashy or smoky, and you bought the coffee

Usually because: The roast went further, or was scorched more, than you like. This was decided before the bag was sealed.

Change first: Nothing in brewing removes it. A coarser grind or cooler water softens it a little.

Harsh or burnt taste

Grassy, sharp or raw, and you bought the coffee

Usually because: Often a very light roast, sometimes a faulty one. Check first that the brew is not simply under-extracted.

Change first: Brew it finer or hotter once. If it is still raw, the cause is the roast, not you.

Grassy or flat, and you roasted it yourself

Usually because: The roast ended earlier, or ran more slowly, than the coffee needed. Which of the two cannot be told from the taste alone.

Change first: Change one thing on the next batch, either the end point or the pace, and keep notes.

Home roast tastes grassy or flat

Dark faces and pale faces in one batch

Usually because: The beans were not mixing through the heat evenly.

Change first: Roast the batch size the maker specifies before changing anything else.

What this page does not claim

CoffeeHQ explanation

It does not give the time, temperature or curve shape that causes any fault. The figures in the table are one laboratory's recipes for producing faults on demand, which is a different thing from the conditions under which they arise by accident.

It does not cover smoke taint from poor ventilation or the effect of slow cooling, because no source on either was opened, and it leaves faults that originate in the green coffee, such as pale beans that will not brown, to the pages on green coffee and grading.

What to do next

• Reading a roast profile — What the second research group found when it tested the curve shapes roasters blame for these faults.
• Buying and judging green coffee — Faults that are in the coffee before it reaches the roaster and only show afterwards.
• Roasting coffee at home — Batch size, supervision and the makers' limits that prevent the worst outcomes.
• Tasting Lab — A worksheet for tasting two roasts side by side and recording what you find in your own words.