efractometers, density meters, agtron readings — there's a lot of equipment in modern roasting, and most of it is genuinely useful. A good roaster's data sheet now has more numbers on it than a small-business tax return. But the most useful tool on the table is still the cupping spoon, and what it tells you tends to arrive a few hours before the data does.
Tom and I cup every batch on Thursday morning, the day after roasting. We pour the samples blind. We grade them blind. Then someone reads out the curve from yesterday and we see whether the cup agreed with the data, or whether the data agreed with the cup. It's almost always the cup that was right first.
The cup is faster than the data because the cup is the only instrument that's already been calibrated for the thing you actually want to know — does this taste good?
A development time of 22% sounds like a good development time. It is, in most contexts. But a 22% development on a washed Yirgacheffe that's been heat-stressed in the drop tastes like burnt sugar wrapped around a vegetable. A 18% development on the same coffee, with a gentler approach to first crack, tastes like the coffee was meant to. The data doesn't tell you which of those happened. The cup does.
That isn't a defence of intuition. It's an observation that the things we measure are only proxies for the thing we actually care about. The numbers exist because cupping is slow and tasting blind is expensive, not because the numbers are smarter than the cuppers. When the numbers and the cup disagree, the cup wins, and we update what we were measuring.
What good cupping actually is
Cupping properly takes about forty minutes. You grind, you pour, you break the crust at four minutes, you skim, you wait until the cups are drinkable, you score in a fixed order, you score blind, and you compare. People who say cupping takes ten minutes are drinking coffee and writing words; cupping is something else.
The thing nobody tells you about cupping is that what you're doing isn't really tasting. What you're doing is comparing. The flavour wheels and the notes and the scores are all ways of compressing one cup so you can hold it against another cup in the same memory slot. The score sheet is a memory aid for the thing you tasted thirty seconds ago.
Cupping isn't really tasting. It's comparing. The score sheet is a memory aid for the thing you tasted thirty seconds ago.
We score every batch on a 1–7 scale on six metrics. Most of the numbers cluster around 5. The interesting data isn't the average — it's the variance. A batch that scores 6/6/6/5/6/6 is a known-good coffee. A batch that scores 7/7/3/6/5/5 is doing something weird, and weird is where you find out what your roaster actually does to a bean.
When the printout wins
To be fair, there are times when the data catches things the cup misses. Quaker detection — under-developed beans that look fine, score fine, then turn paper-flavoured in espresso a week later — is one. Defect density at the sorting stage is another. Anything statistical, anything that requires you to keep score across a hundred batches, the data wins. Anything that requires you to remember whether a coffee tasted right, the cup wins.
The mistake most new roasters make is to think they have to pick one. You don't. You collect the data, you cup the coffee, and when they disagree you cup it again. The coffee gets better when you take both seriously. It gets worse the moment you trust one and stop questioning the other.
— Iris Chen
Head Roaster, Bramble Coffee