How a roaster builds a blend

A blend is a specification, not a mixture

The useful way to think about a house blend is as a brief. A roaster writes down what the cup must do: sweet, chocolate-forward, holds up in milk, 88 to 91 °C brew window, works at 1:2 in eighteen seconds, and tastes the same in March as in October.

Everything else follows from that brief. The countries are interchangeable. The brief is not.

This is exactly backwards from how a single origin works, where the place comes first and the cup is whatever the place gave you that year. It is also why a good blend is harder to make than a good single origin, and why a bad one is so much worse.

The three roles

Almost every blend, whatever it is called, is built out of three jobs.

The base. Sixty to eighty per cent, usually. It carries body, sweetness and the low-frequency chocolate and nut. Brazil dominates here for good reasons: natural processing, low acidity, soft sweetness, enormous and reliable supply at a workable price. Peru, Honduras and Indian arabica do the same job.

The middle. Ten to thirty per cent. This is structure: a washed Central American, usually, adding clean acidity and a defined shape so the base does not read as a flat brown wall. Colombia, Guatemala, Honduras washed, El Salvador.

The top. Five to fifteen per cent, sometimes less. This is the note you remember: an Ethiopian for florals and citrus, a Kenyan for blackcurrant, a natural for fruit and ferment. It is the smallest and most expensive component, and it is the first one a roaster cuts when green prices move.

In an espresso blend there is often a fourth role, which is grip: a fine robusta or a heavy Indian arabica for crema and persistence through milk.

The mistake amateurs make when blending at home is using three coffees that all want to be the top. Three interesting coffees make one confusing cup.

Pre-blend or post-blend

This is the choice that separates a workshop from a factory, and both answers are defensible.

Pre-blending, or blending before the roast, means the green components go into the drum together and come out as one roast. It is one roast instead of four, one batch to cool, one silo, one lot to track. Cheaper, faster, less to go wrong in a small shop.

The catch is density. A dense washed Colombian from 1,900 metres and a soft Brazilian natural from 900 absorb heat at different rates and reach first crack at different moments. Roast them together and one is under-developed while the other is baked. The usual mitigation is to blend only components of similar density and screen size, which quietly constrains the recipe.

Post-blending, or blending after the roast, means each component gets its own profile and the roasted coffees are weighed together afterwards. The Ethiopian can stay light enough to keep its florals while the Brazilian goes far enough to develop its sugars.

The cost is real: four roasts instead of one, four cooling cycles, four bins, four chances to weigh wrong, and a minimum batch size per component that a small roaster may not be able to sell through before it goes stale.

Most specialty roasters post-blend their flagship and pre-blend their volume lines, which tells you everything about where the extra cost is worth paying.

How the recipe is actually found

Not by theory. By cupping, in three passes.

First, the components are cupped alone at a standard profile, and the roaster writes down what each one contributes and what it costs. Second, pairs are cupped against each other, because most blend failures are one pair fighting rather than a whole recipe being wrong. Third, candidate ratios are cupped side by side, usually three or four variants a few percentage points apart.

Then, and this is the step that gets skipped, the shortlist is brewed the way customers will actually brew it. A blend that cups beautifully at 1:18 in a cupping bowl can fall apart as an 18-gram espresso at 1:2, and the espresso is what the bag says it is for.

A ratio that survives all four passes gets written down as a recipe: percentages by green weight, a profile per component, and a cup description precise enough to audit against later.

The part nobody sees: holding it still

Harvests arrive at different times of year and last a few months in the warehouse. The Ethiopian that made the blend sing in April will be gone by November. The Brazilian will be a different lot from a different farm with a different screen size.

So the recipe is rewritten, typically two to four times a year, and the target is not the old recipe. The target is the old cup.

The workflow is the same every time: cup the incoming lot against the outgoing one, adjust the percentages, adjust the profile, cup the new blend against a retained sample of the previous batch, and accept it only when a taster cannot reliably tell them apart. A good roaster keeps a frozen reference sample of the blend as it should be, for exactly this.

This is why blend components are so rarely printed on a bag, and why a roaster who does print them usually adds a harvest date beside each one.

What it costs, and where the money goes

Green coffee is bought at very different prices. In a typical bag, the base costs the least per kilogram and occupies most of the weight; the top costs three to ten times as much and occupies a tenth of it. A blend is therefore a cost engineering exercise as well as a flavour one, and shifting the top component from 12 per cent to 8 per cent is one of the quietest ways a roaster can absorb a price rise.

That is not cynicism. It is the mechanism, and knowing it is how you read a blend that has drifted. When a bag you have bought for two years suddenly tastes flatter and less aromatic without tasting wrong, the top note is usually what left.

Blending at home, which is worth doing once

You cannot pre-blend without a roaster, but you can post-blend with a scale, and an afternoon of it teaches more about blending than any amount of reading.

Buy three roasted coffees that are deliberately unalike: a Brazilian natural, a washed Colombian or Guatemalan, and a small bag of something Ethiopian. Brew each one alone first and write down one sentence about what it gives you. Then weigh blends at 70/20/10, 60/30/10 and 60/25/15 by bean weight, grind each fresh, and brew all three the same way.

Two things usually happen. The first is that the Ethiopian, which you expected to dominate at 10 per cent, does not: aromatics are fragile and the base swallows them. The second is that one of the three blends is clearly rounder than any of the single coffees was, and you will not be able to say why from the ingredient list alone. That is the whole job, felt rather than explained.

One caution. Grind and brew the components together, not separately into the same cup. A blend is an extraction as well as a recipe, and beans of different density extract at different rates in the same bed, which is a real part of what you are tasting.

How to judge a blend you are buying

• Does it name a purpose? A blend built for espresso with milk and a blend built for filter are different recipes. A bag that claims to be for everything is usually for nothing in particular.
• Is there a roast date? A blend of four components staling at four different rates needs freshness more than a single origin does, not less.
• Does it hold up across the year? Buy the same bag in March and in September. A house blend that tastes noticeably different is not being maintained, and maintenance is most of the craft.
• Does it resolve? The test of a blend is not how many things you can taste in it. It is whether the things you taste arrive as one cup rather than as three coffees taking turns.