How cocoa quality and flavour are professionally assessed
A buyer, an exporter and a competition judge can only compare notes if they sampled, roasted, ground and tasted the same way. Since 2023 there has been an internationally agreed way to do it. This page sets out what that protocol specifies, step by step, and what each kind of result can and cannot tell you.
Read this first
- The protocol is voluntary. It is a common yardstick, not a legal standard and not the only sound method.
- It assesses cocoa beans. It is not a method for judging a chocolate bar bought in a shop.
- This site’s own tasting scorecard and tasting flight are for people tasting finished chocolate for themselves. They use a different vocabulary and scale, and a score from them is not comparable with a panel’s.
- The guide’s forms and flavour wheel are not reproduced here. They are published by their authors under a non-commercial licence; the sources below link to the original.
Five things called quality
A lot can pass one of these and fail the next. Most confusion about cocoa quality comes from treating them as one.
| Kind of quality | How it is assessed | What it tells you | What it does not |
|---|---|---|---|
| Physical bean quality | Moisture content, bean count and size, sieving, and the cut test on 300 beans | Whether a lot is sound: dry enough, even in size, free of mould, insects and germination | How it tastes. A physically perfect lot can be flat or harsh. |
| Degree of fermentation | The colour and fissuring of cut beans, and their aroma | Whether beans are slaty, partly fermented, well fermented or over-fermented | Flavour directly. Colour is a proxy, and pale-seeded types do not turn the same brown. |
| Sensory quality of the beans | A trained panel tasting the beans as unroasted coarse powder and as roasted cacao mass | The intensity of each flavour attribute and any off-flavour, on a 0 to 10 scale | What a maker's own roast and recipe would produce. |
| Flavour potential | The same beans made into a standard 70% dark chocolate and tasted | How the lot expresses itself in chocolate made one fixed way | The best it could be. One standard recipe is a common yardstick, not an optimum. |
| Quality of a finished chocolate | Outside this protocol, which assesses beans | Nothing here: a commercial bar reflects its recipe, conching and age as well as its beans | The quality of the beans it was made from. |
From sack to score
The steps in order, each with the figure the protocol gives. The sections below quote the passages.
| Step | What the protocol specifies |
|---|---|
| Sample the lot | 30% of bags; at least 300 g per tonne; a reference sample of at least 2,000 g |
| Measure moisture | On a test sample of at least 500 g |
| Assess whole beans | Sieving, cleaning loss and bean count, each on at least 500 g |
| Cut test | 300 beans: defects counted, colour and fissuring graded, aroma noted |
| Roast | At least 600 g in a forced-air convection oven: 112, 120 or 130°C for 25 minutes, adjusted for bean size and moisture |
| Break and winnow | To no visible shell |
| Refine to cacao mass | To 14 to 20 micrometres, never above 55°C |
| Make the standard chocolate | 63% mass, 7% cocoa butter, 30% sugar; 18 micrometres or finer; never above 55°C |
| Taste the mass | 1 to 2 g, melted at 48 to 50°C for no more than five minutes; coded, in random order; at most six in a session after two reference samples |
| Taste the chocolate | 2 to 3 g at room temperature |
| Score | Each attribute 0 to 10 for intensity, then a global quality score 0 to 10; at least six assessors for an averaged profile, four for a consensus |
These are the guide’s key specifications, not a working method. Anyone setting up a laboratory or a panel needs the guide itself, with its equipment lists, adjustment tables and forms.
What this protocol is, and its standing
Cocoa was graded for centuries by eye and by each buyer's own tasting. An internationally agreed method is recent.
- The guide was published in 2023. Its introduction says that until then there were no commonly agreed protocols for assessing cacao quality and flavour. (source 1)
- It is voluntary. Nobody is obliged to use it, contracts are not governed by it, and a buyer's own laboratory may roast and taste differently. (source 1)
- It assesses beans in three forms: unroasted as a coarse powder, roasted and ground to cacao mass, and made into a standard dark chocolate. (source 1)
Not shown: How widely the protocol is used. No figure for adoption by buyers, exporters or national bodies was found.
Not shown: How it relates to the grading clauses of trade contracts, which were not read. The guide itself notes that commercial arbitration has its own cut-test requirements.
Not shown: Any later edition. The 2023 text is the one read.
Physical evaluation and the cut test
- A lot is sampled before anything else: 30% of the bags, at least 300 g of beans per tonne, combined into a reference sample of at least two kilograms. (source 1)
- The cut test is done on 300 beans, cut lengthways. The guide notes that companies tend to start with 50 and stop if those show no defect. (source 1)
- Defective beans are counted in four classes, in descending order of severity: mouldy, insect-damaged, germinated and slaty. A bean with two defects counts under the worse one. (source 1)
- The cut beans are smelled as well as looked at, for smoky, mouldy or hammy odours. That first impression decides how the sample will be roasted. (source 1)
Not shown: The grade limits themselves. How many defective beans a grade allows is set by each country's standard and each contract; the guide tabulates several in an annex that was not transcribed here.
Not shown: That a clean cut test means good flavour. It screens out damage and shows the degree of fermentation, and the guide treats flavour separately.
Preparing the sample: one way, so that results compare
Most disagreement between tasters starts before anyone tastes. The protocol fixes the roast, the grind and the recipe so that a difference in the cup is a difference in the beans.
- Three basic roasts are defined for a laboratory convection oven: 112, 120 or 130°C, each for 25 minutes. The time and temperature are then adjusted by a table for the size and moisture of the beans. (source 1)
- Which roast is used depends on the aromas found in the cut test and the dominant genetic type, where it is known. The protocol does not roast every cocoa alike. (source 1)
- The roasted nibs are refined to a mass of 14 to 20 micrometres and kept below 55°C throughout, so that grinding adds no further roast. (source 1)
- For the chocolate stage every sample is made to one recipe: 63% mass, 7% cocoa butter and 30% sugar, with nothing else. (source 1)
Not shown: That these roasts are the best for any particular cocoa. They are a standard for comparison. A maker's own roast is a different decision with a different purpose.
Not shown: Conching time and tempering details, which the guide describes and which were not summarised here.
How the tasting is run
- Cacao mass is served melted at 48 to 50°C, one or two grams at a time, and is not held warm for more than five minutes or reheated. Chocolate is tasted at room temperature. (source 1)
- Samples carry random three-digit codes and are served in a randomised order, so that nobody knows what they are tasting or tastes in the same sequence as their neighbour. (source 1)
- Between samples assessors rinse with warm water and chew a plain cracker made only of flour and water. (source 1)
- Sessions are short. With reference samples tasted first for calibration, the guide advises a maximum of ten samples and a rest before the next session. (source 1)
- A panel needs at least six trained assessors if the profile is the average of their scores, and at least four if it is agreed by discussion. The guide says consensus profiles reproduce less well across panels and over time. (source 1)
Not shown: How assessors are selected, trained and monitored in detail. The guide has a chapter on it, including statistical checks of each assessor, which is not summarised here.
Not shown: How closely two panels following the protocol agree. No inter-laboratory comparison was read.
What is scored, and on what scale
- Each attribute is scored for intensity from 0 to 10, and the numbers have stated meanings: 0 is absent, 3 to 5 clearly characterises the sample, 6 to 8 dominates it. It is a scale of how much, not of how good. (source 1)
- The attributes are in three groups. Core attributes are expected in every sample: cacao, acidity, bitterness and astringency, with roast degree. Complementary attributes may or may not appear. Off-flavours come from defects. (source 1)
- The complementary attributes are fresh fruit, browned fruit, vegetal, floral, woody, spice and nutty. Sweetness is scored only for chocolate. (source 1)
- Quality is a separate, single score from 0 to 10 given after the profile. It rewards the absence of off-flavours, balance among the core attributes and outstanding complementary ones. (source 1)
Not shown: Reference standards for each attribute. The guide's glossary ties intensity levels to example origins, which is its authors' experience and is not repeated here as a statement about those origins.
Not shown: That a high global score predicts a price or a prize. The score is one panel's judgement by one method.
Off-flavours: what counts as a defect
- The off-flavours named are dirty or dusty, musty, mouldy, meaty or animal or leather, over-fermented or rotten fruit, putrid or manure, and smoky, with a class for others such as rancid, diesel and chemical taints. (source 1)
- An off-flavour is tolerated as a trace. Scored at 3 or more, it characterises the sample as defective. (source 1)
- Smoke is a contaminant in this scheme, whatever its source. A cocoa valued locally for a smoky note would be marked down by this protocol. (source 1)
Not shown: The cause of any off-flavour in a given lot. The protocol detects; it does not diagnose. The fermentation and drying pages cover causes.
Not shown: Chemical confirmation. A sensory panel is not a residue or mould-toxin test, which are separate methods.
Sources
Studies are cited by author and year and summarised, with their limits, on the research pages. The official documents are listed here. Each was downloaded and read, and the passage relied on was checked to be in it.
- Bioversity International. Guide for the Assessment of Cacao Quality and Flavour (Cacao of Excellence, Alliance of Bioversity International and CIAT, with the working group on International Standards for the Assessment of Cocoa Quality and Flavour) (2023-09-26). hdl.handle.net/10568/132065
Voluntary international text · International
Reviewed 9 October 2026.
Related: testing methods, assessing cocoa bean quality, fermentation, the fermentation studies, roasting, the genetics studies, flavour vocabulary and how to taste chocolate. All data.