Blend Designer

Think of blending as "hiding weak beans behind strong ones" and the design goes wrong from the start. The essence of a good blend is creating a direction of flavor no single component can reach alone. Layer Brazil’s nuttiness under Ethiopia’s florals and you gain the body Ethiopia lacked on its own — while the Brazil gains the lift it never had. One plus one makes three; that is the craft.

A blend also hedges dependence on any single origin. In a year when one country’s harvest disappoints, a multi-origin blend barely moves. Espresso blends are usually built from 3–5 origins exactly for this: stability of flavor and economics secured at once.

One of the oldest blends on record is "Mocha Mattari" — beans of several origins shipped together from Yemen’s port of Mokha, which later generations came to call a single name. In 19th-century Europe, the standard was the "Mocha-Harrar" type: inexpensive South American coffee lifted with a little Yemen or Ethiopia.

The modern blend matured in 20th-century Italy. With espresso culture came the grammar of origin roles: Brazil for the base, Colombia for sweetness, Robusta for crema, Ethiopia for aroma. Italian espresso blends still mostly follow that basic structure today.

Japan had its own golden age in the kissaten blends of the 1970s–80s. Two-bean standards easy to reproduce at home — Mandheling and Mocha, Colombia and Brazil — were born then, and their DNA survives in the designs of today’s third-wave roasters.

As an introduction to blend design, the three-layer structure works best: base 50–60%, accent 20–30%, body 10–20%. The base is a mild, low-quirk bean (Brazil, Honduras); the accent is a strongly characterful bean (Ethiopia, Kenya); the body layer carries weight and sweetness (Guatemala, Sumatra).

Tilt the ratio toward one axis — say 70-20-10 — and the character stands up; even it out to 50-30-20 and you get a balanced, "legible" cup. Move the sliders while watching the radar chart and you can see which axis grows as you raise each part.

Beginners are safest starting from proven pairs. Brazil × Ethiopia is the all-purpose pair (nutty × floral); Colombia × Kenya the vivid one (sweetness × berry); Guatemala × Brazil the dark-roast-friendly balance (chocolate × nut). Conversely, two high-acid or two big-personality beans tend to fight, and the design gets difficult.

• Base at the middle of the roast range, accents at the edges — an easy rule of thumb
• Mix like roast levels with like. Deep × light tends to separate into clumps of aroma
• Start with 2 components; expand to 3, then 4 as you get comfortable

There are two broad ways to blend: mixing green and roasting together ("pre-blend" / melange roasting), or roasting separately and mixing after ("post-blend"). The same recipe tastes meaningfully different depending on the order.

Pre-blending puts every bean through the same heat history, so the flavors fuse into a smooth, unified cup — at the cost of compromise when components really want different roast levels. Post-blending roasts each bean to its own optimum before mixing, maximizing each one’s character, though the flavors can stand slightly apart.

Professionals often say post-blending is easier to quality-control. For home enjoyment, the realistic method is exactly that: buy two or three roasted coffees and post-blend them yourself. This simulator is designed around that assumption.

The simulator predicts via the radar chart, but the real cup drifts from prediction with grind, water temperature, brew time and roast variance. The shortcut to getting good is always brewing the blend you designed and recording where reality diverged.

For records, this site’s tasting log is the tool: it can display predicted vs measured side by side on the radar chart, making the gap between design and perception visible. Brew the same blend about three times and compare, and adjustment guidance appears — "this balance is too forward for my palate," "the base wants a little more."

A good blend rarely lands on the first try. Simulate → drink → adjust, three to five times, and you finally reach the recipe that clicks. Use the simulator as the map that cuts down those iterations.