Roasting Cacao: A Profile Design Guide by Origin

Roasting is the step every new bean-to-bar maker obsesses over and the step most experienced makers eventually realize they were over-thinking. The roast doesn't make your chocolate — fermentation does that. But the roast decides whether your fermented cacao reaches its potential or gets flattened into generic “chocolate flavor.” This post is the working guide to cacao roast profile design: what the roast actually does, the three dimensions of a profile, the starting ranges by origin, and the three-profile ladder method every careful maker uses to dial in a new lot.

If you haven't read our fermentation primer yet, start there — the rest of this post assumes you understand that fermentation creates the flavor precursors, and roasting converts those precursors into finished aroma. You can't profile-design what you can't diagnose.

What roasting actually does

Cacao roasting performs four distinct functions, and every profile is a balance between how far you push each one:

• Maillard reactions.The free amino acids and reducing sugars produced during fermentation react under heat to form pyrazines, aldehydes, and other chocolate-characteristic aromatics. This is where “chocolate flavor” is created. Not enough time above 100°C and you get raw-bean character; too much and Maillard runs past the origin notes into generic cocoa.
• Acetic acid removal. Fermentation deposits volatile acetic acid (vinegar) inside the bean. Roasting drives it off. Under-roasted beans taste sharp and sour; over-roasted beans have lost the brightness acid would have preserved.
• Moisture reduction. Raw fermented beans are 6.5–7.5% moisture. A good roast drops them to 1.5–2.5% — dry enough to crack cleanly through a winnower and stable enough that downstream grinding produces a workable chocolate.
• Husk loosening. Heat separates the papery husk from the nib so winnowing yields cleanly. Beans that were roasted too cool or too briefly are miserable to winnow, which tanks yield and drops margin.

The three dimensions of a roast profile

Any roast profile can be described in three numbers. Getting the right combination is the whole game.

1. Peak bean temperature

The highest temperature the bean mass reaches. Measured at the bean itself, not the air inside the roaster — this is a common rookie mistake. Use a probe thermometer or a sampling tool that lets you grab and check beans during the roast. Typical range for craft bean-to-bar: 110–140°C (230–285°F).

2. Total roast time

From the moment beans enter the hot roaster to the moment they leave it (ideally into a forced-air cooler). Typical range: 12–35 minutes. Shorter roasts preserve top-note character at the risk of leaving residual acid; longer roasts drive off more acid but risk flattening flavor.

3. Rate of rise

How quickly the bean temperature climbs through the mid-roast range (roughly 90–130°C). Too fast (>4°C/min) produces scorched edges with still-raw centers. Too slow (<1°C/min) is a “baked” profile: flat, muted, papery. Target for most craft roasts: ~2–3°C/min through the middle, slowing as you approach peak temperature.

Reading a roast curve

If your roaster lets you log bean temperature over time — or even if you just sample and note temperature every 60 seconds — you can plot a roast curve and read it like a signature. A healthy craft-cacao roast has three phases:

1. Drying phase (0–6 min) — the bean surface temperature climbs quickly from ambient toward ~100°C while moisture evaporates. Bean mass temperature rises slowly because the evaporation is pulling heat.
2. Development phase (6–18 min) — once moisture is mostly gone, the bean temperature climbs steadily through the Maillard-reaction window (~110–130°C). This is where flavor is built. The rate of rise during this phase is the most consequential number in your profile.
3. Peak and hold (18–25 min)— the bean reaches its target peak and holds briefly before the roaster heat is cut. The audible sound of husk cracking often indicates you're entering this phase; it's also a signal many roasters use as a rough peak-temperature indicator.

Origin-specific starting ranges

Every lot is its own data point, so treat these as starting profiles — places to begin iterating, not gospel. Conservative, delicate origins (Madagascar, Ecuadorian Nacional, high-end Venezuelan) want cooler, shorter roasts. Robust, bold-flavored origins (West African, some Vietnamese, Tanzanian) tolerate and often benefit from hotter, longer ones.

The three-profile ladder method

Experienced makers almost never hit the right profile on the first try with a new lot. The three-profile ladder is the standard iteration protocol. It takes a day, costs ~1.5 kg of cacao, and prevents you from committing a full 25-kg lot to a bad profile.

1. Sample-roast 500 g at three profiles. Low/slow: target the cool end of the origin range (e.g. 115°C, 22 min for a Madagascar). Medium: middle of the range (e.g. 120°C, 20 min). High/fast: the warm end (e.g. 125°C, 17 min).
2. Winnow and taste the nib from each. You're not making chocolate yet — just evaluating the roasted nib. Does the low/slow taste raw and acidic? Does the high/fast taste flat? Is one of the three clearly more balanced?
3. Pick a direction for the next ladder. If the low/slow was best but still slightly raw, your production profile lives between low and medium. If the medium was best, run a narrower ladder around medium. Three ladder iterations usually converges on a production profile.
4. Confirm with a 500 g test batch. Once you have a candidate profile, roast 500 g, refine into a simple 70% recipe, and taste after conching. The cacao behaves differently as finished chocolate than as nib — confirm before committing the full lot.

Equipment-specific considerations

Drum roasters (Behmor, Diedrich, small coffee drums)

The most common setup for small bean-to-bar production. The drum's tumbling action gives even roasting and easy sampling. The main adjustment vs. coffee: cacao roasts dramatically cooler than coffee. If you're converting a coffee roaster, don't use coffee's 190–230°C profiles — you'll scorch the beans. Cacao caps out around 135°C for most origins.

Convection ovens

A reasonable hobby / pre-production setup. Spread beans on a perforated tray in a single layer; stir every 3–4 minutes to prevent uneven roasting. Convection ovens lie about their temperatures — use an independent oven thermometer. Most home convection ovens are 10–15°C cooler than their dial suggests, which is why so many home-roasted batches are actually under-roasted.

Hot-air poppers (popcorn machines)

The classic entry-level option for testing ideas. Small batch sizes (200–300 g), fast roasts, very loud. Poppers heat fast and don't easily sustain long development phases, so they tend to produce shorter, hotter profiles than is ideal for delicate origins. Fine for testing robust origins; not recommended for Madagascar or Nacional lots where subtlety matters.

Dedicated cacao roasters

Professional machines like the Cacao Cucina, Selmi, or custom-built infrared roasters designed specifically for cacao. If you're producing 1,000+ bars a week you'll eventually want one. Before that, drum roasters with careful profiling will produce chocolate that competes at international award levels.

Documenting a profile so you can reproduce it

A profile you can't reproduce is a lucky accident. The minimum log for each roast should include:

• Cacao lot ID and origin. Same bean, same cooperative, same harvest year — never conflate lots.
• Green bean weight and moisture. Input weight and starting moisture (from CoA or your own test) matter because they affect how the profile develops.
• Ambient conditions. Room temperature and humidity on roast day. Humid days slow drying; cold days slow ramp.
• Roaster settings. Charge temp (air temp when beans enter), ramp profile (heat setting changes through the roast), fan setting.
• Key time/temperature waypoints. Bean temp every 60–90 seconds; audible crack timing; sample-pull observations.
• End conditions. Peak bean temp, total roast time, post-roast cooling time, output weight (for yield tracking).
• Tasting notes from finished chocolate. Tie the roast back to the finished bar's sensory profile. A roast without a taste log is a number without meaning.

Common roast mistakes and how to hear them

The seasonal drift problem

A profile that worked last year may not work this year even with the same cooperative. Cacao is an agricultural product; weather, fermentation variation, and farmer rotation all shift the bean's properties. Plan to re-profile any time you receive a new lot, even from a known supplier. A good maker tastes the raw nib before roasting and adjusts — under-fermented cut tests want a longer, gentler roast to develop Maillard; well-fermented lots can handle shorter, hotter profiles because the precursors are already concentrated. We go deeper on reading those cut tests in our fermentation guide.

The cheat sheet

Roasting is the bean-to-bar step that rewards patience and documentation more than any other. The maker who profiles every new lot with a ladder, writes down the settings, and ties the roast back to finished-bar tasting notes is building an institutional memory their chocolate compounds on. The maker who eyeballs it and doesn't write anything down is essentially starting from scratch every harvest.

For the broader production context, see our seven stages of bean-to-bar production guide — roasting is stage 2, and each downstream stage (winnowing, grinding, conching, tempering) interacts with how you profiled. For sourcing considerations that affect which profiles are possible, read our direct-trade sourcing guide.