ChocolateHQ

How cocoa and chocolate are tested

“Lab tested” covers a great many different things. Here is each method in use, what it actually measures, and what a result from it does not tell you.

17 methods5 purposesEach with a published use

How to read an entry

Every method answers one narrow question. Each entry says what that question is, gives one or two studies that used the method and what they found, and then lists what such a result does not establish. Studies in this site’s research register are linked by author and year; other papers are numbered and listed at the foot. All were read as published abstracts unless marked otherwise.

This is a description of what methods are for. It is not a set of procedures, and nothing here is a specification for a laboratory.

Raw beans

Judging a lot of fermented, dried beans before anything is made from it.

Cut test

Also: visual inspection, bean cut test

A sample of beans is cut lengthways and each is judged by eye, chiefly on the colour and texture of the cut face, and counted into classes such as well fermented, under-fermented, mouldy or insect-damaged.

In use

  • A 2015 paper describes quality control of fermented cocoa as being achieved, to date, by visual inspection such as the cut test and by sensory testing. (source 1)
  • It is used as an outcome measure in process trials. In one, 49% of beans were classed as well fermented with part of the pulp removed and 48% without. (Streule et al. 2024)
  • It can miss what a taster finds. Beans dried at 80 °C and at 60 °C showed no difference by cut test or fermentation index, and the hotter-dried ones gave a more bitter and astringent liquor. (source 14)

What a result does not show

Flavour. It classifies appearance, and two lots with the same cut-test result can taste different.

Anything about a lot beyond the beans cut. It is a sample, and the classes are a judgement by eye.

Fermentation index

Also: FI

A laboratory figure, read from how an extract of the bean absorbs light, that rises as fermentation proceeds.

In use

  • In box fermentations of Mexican cocoa sampled every 24 hours, the value suggested as the minimum adequate, 1 or above, was reached at 72 hours, at which point pH, protein breakdown and other measures also indicated well-fermented beans. The authors propose pH as a first indicator during fermentation, confirmed by the index. (source 3)
  • Predicting it for single beans by hyperspectral imaging worked only moderately, explaining about half the variation in validation. (source 2)

What a result does not show

That the fermentation was a good one. It tracks one consequence of fermentation among several.

A universal threshold. The figure of 1 is a suggested minimum reported for particular beans.

pH and acidity

Also: titratable acidity, volatile acidity

How acid the bean or the fermenting mass is. pH is the concentration of free acid; titratable acidity is the total amount of acid present; volatile acids such as acetic can be measured separately.

In use

  • Drying method changed bean acidity and volatile fatty acids, and the chocolate with it: oven-dried beans kept high volatile acid and gave a strong off-flavour. (Jinap, Thien and Yap 1994)
  • pH was proposed as a practical first indicator of the end of fermentation, usable on undried beans during the process. (source 3)
  • One study of over-fermentation suggested measuring the pH of the fermenting mass as an alternative to cutting beans daily. (source 15)

What a result does not show

Which acids are present. A vinegar-like and a yoghurt-like acidity can give the same pH and taste quite different.

Near-infrared spectroscopy

Also: NIR, NIRS, hyperspectral imaging

How a sample absorbs near-infrared light. The spectrum means nothing by itself: a statistical model, built beforehand from samples analysed the slow way, converts it into a predicted value.

In use

  • A model predicted phenolic substances, organic acids, epicatechin, lactic acid, fermentation time and pH in fermented cocoa, each with a coefficient of determination between 0.87 and 0.94. (source 1)
  • Imaging single beans, models on 170 beans from 17 batches predicted fermentation index, total polyphenols and antioxidant activity with validation coefficients of 0.50, 0.70 and 0.74, which the authors call sufficient for screening. (source 2)

What a result does not show

Anything outside what its model was trained on. A calibration built on one set of origins or one season is not thereby valid for another.

A measurement in the strict sense. It is a prediction, and the single-bean results show how wide the error can be.

Flavour and process

Finding out which compounds are present and which of them matter.

Liquid chromatography

Also: HPLC, UHPLC, LC-MS

Separates the non-volatile compounds in an extract and quantifies them one by one: the flavanols, the larger polyphenols, theobromine and caffeine, sugars, acids.

In use

  • It is how the loss of catechin and epicatechin through fermentation, drying, roasting and alkalising was quantified, with fermentation removing most. (Payne et al. 2010)
  • It tracked epicatechin and theobromine falling during heap fermentation, and tied what remained to how bitter the chocolate was. (Camu et al. 2008)
  • Coupled to high-resolution mass spectrometry, it identified polyphenol markers that separated fermented from unfermented beans and beans of different origins. (source 6)

What a result does not show

What the compounds taste like in the chocolate. That needs tasting, compound by compound.

A health effect. A flavanol figure describes the product, not what it does in a body.

Gas chromatography with mass spectrometry

Also: GC-MS, headspace GC-MS, volatile profiling

Separates the volatile compounds given off by a sample and identifies each from its mass spectrum. A sample of cocoa yields dozens to hundreds.

In use

  • Fifty-eight volatiles were identified across fermentation times and drying temperatures to compare treatments. (Rodriguez-Campos et al. 2012)
  • For smoke taint it has been turned into a routine check: ten marker compounds were quantified and ranges set for accepting or rejecting incoming beans, the markers being 7 to 125 times more concentrated in smoky samples. (source 10)

What a result does not show

Which of the compounds found can be smelled. Most volatiles in cocoa are below the concentration at which a nose detects them.

Flavour quality. A long list of volatiles is an inventory.

Gas chromatography with olfactometry

Also: GC-O, aroma extract dilution analysis, AEDA

The same separation, with a trained person smelling each compound as it leaves the instrument. Repeating it on ever more dilute extracts ranks the compounds by how potent an odorant each is.

In use

  • It identified 33 potent odorants in dark chocolate and found that conching creates no new key odorant, only changing amounts. (Counet et al. 2002)
  • Its strongest form of proof is reconstruction: mixing 24 identified odorants at their measured concentrations reproduced the typical smell of the cocoa powder they came from. (source 11)

What a result does not show

How the chocolate tastes as a whole. It ranks odorants one at a time, outside the fat and sugar they are eaten in.

Finished chocolate

Flow, structure, texture and how it is perceived.

Rheometry

Also: viscometry, Casson yield value, plastic viscosity

How molten chocolate responds to being sheared at a controlled rate or stress. Two things are read from it: the stress needed to start it moving, and its resistance once it is flowing.

In use

  • It showed fat content, particle size distribution and lecithin each changing every flow property of dark chocolate, with fat the largest influence. (Afoakwa, Paterson and Fowler 2008)
  • The widely used Casson model fits the data well between shear rates of 5 and 60 per second and cannot resolve real differences below 5; an oscillatory measurement separated chocolates the Casson fit could not. (De Graef et al. 2011)

What a result does not show

Whether a chocolate will mould, enrobe or pan well. The numbers describe flow under the instrument's conditions; what a given job needs is trade practice, not something these studies establish.

A single 'viscosity'. Chocolate has no one value; it depends on how fast it is sheared.

Particle size distribution

Also: PSD, D90, fineness

The sizes of the solid particles suspended in the fat, reported as a distribution. D90 is the size below which 90% of the particles fall.

In use

  • It is the variable set in studies of flow and texture: chocolates with D90 of 18, 25, 35 and 50 micrometres differed in viscosity, hardness, colour and gloss. (Afoakwa, Paterson and Fowler 2008; Afoakwa et al. 2008)
  • The shape of the distribution matters as well as its top end: roller-refined and ball-milled chocolates differed in the spread of sizes and, with it, in flow. (Bolenz and Manske 2013)

What a result does not show

Smoothness as a taster experiences it, from one number. People differ in what they can feel, and a D90 says nothing about the rest of the distribution.

Differential scanning calorimetry

Also: DSC, melting profile

The heat a small sample absorbs as it is warmed. The temperatures at which melting starts, peaks and ends, and the total heat taken up, reflect which fat crystals are present and how many.

In use

  • It distinguished optimally tempered, over-tempered and under-tempered dark chocolate: temper changed where melting ended and how much heat it took, while particle size did not change crystallinity. (source 12)
  • With microscopy, it was used to follow how seeding and individual ingredients change the speed of cocoa butter crystallisation. (Svanberg et al. 2011)

What a result does not show

The crystal structure itself. It infers the state of the fat from how it melts.

Trained sensory panel

Also: descriptive analysis, temporal dominance of sensations, TDS

People, trained to use agreed terms on agreed scales, score coded samples. It is the only method here that measures flavour, as opposed to something expected to predict it.

In use

  • It is the final test in the process studies: heaps with the same microbes gave chocolates a panel could tell apart. (Camu et al. 2008)
  • Panels are instruments with their own error. When different trained panels described the same chocolates over time, they gave different descriptions. (source 13)

What a result does not show

What consumers will prefer. A trained panel describes; it is deliberately not asked what it likes.

A result independent of the panel. Vocabulary, training and the number of assessors all move the answer.

Authenticity and origin

Testing whether a product is what it is said to be.

Triglyceride profiling

Also: TAG analysis, cocoa butter purity test

The proportions of the main fat molecules in a sample, by gas chromatography. Cocoa butter has a characteristic pattern; other vegetable fats, including those made to imitate it, have different ones.

In use

  • The five main triglycerides were enough to group genuine cocoa butters of different origins, varieties and seasons apart from cocoa butter equivalents of different types, information the authors say can be used to assess a sample's purity and origin. (source 4)

What a result does not show

From this study alone, how little added fat can be detected in a finished chocolate. That is a question of detection limits in a mixture, which the abstract does not report.

Whether an added fat is permitted. Several jurisdictions allow up to 5% of certain vegetable fats; the standards table on this site gives each rule.

Stable isotope ratios

Also: isotope fingerprinting, IRMS

The ratio of heavy to light forms of elements such as carbon and nitrogen in the bean. The ratios vary with the conditions a plant grew in, so they carry a trace of where that was.

In use

  • Sixty-one samples of fermented beans from 24 origins on four continents could be distinguished by origin using carbon and nitrogen ratios in several tissues and extracts, and the ratios correlated with altitude and rainfall. (source 5)

What a result does not show

The origin of an unknown sample with certainty. Sixty-one samples across 24 origins is two or three per origin; telling groups apart in a study is far easier than assigning one new sample to a place.

Anything once beans are blended, or after seasons the reference samples do not cover.

DNA genotyping

Also: DNA fingerprinting, SSR markers, SNP markers

Which variants a tree or a bean carries at a chosen set of positions in its genome. Comparing the pattern with reference trees shows what it is, or is not.

In use

  • In germplasm collections it exposed mislabelling on a large scale: a study set aside 289 of 1,241 trees as mislabelled, duplicated or hybrid, and found 100 pairs of differently named trees to be identical. (Motamayor et al. 2008)
  • A review reports several DNA-based methods developed to detect CCN-51 mixed into Nacional beans, working on beans and on cocoa liquor. (source 16)

What a result does not show

Where cacao was grown. A genotype identifies the tree's ancestry, and the same clone is planted in many countries.

Quality or flavour.

From the sources read here, how well it works on finished chocolate, where the DNA has been through roasting and conching.

Chemical fingerprinting

Also: metabolomic fingerprint, chemometrics

Not one compound but the whole pattern of many, compared statistically between groups of samples to find the features that differ.

In use

  • A large set of beans yielded polyphenol markers distinguishing fermented from unfermented beans and beans of different origins. (source 6)

What a result does not show

That the pattern reflects origin and not something that travels with it: the variety grown there, the local fermentation practice, or the season. The same study found fermentation status separable by the same means.

Contaminants and allergens

Looking for what should not be there.

Allergen immunoassay

Also: ELISA, allergen test kit

Antibodies that bind a particular food protein, such as one from peanut or milk, give a signal in proportion to how much is present in an extract of the sample.

In use

  • A survey of 92 chocolate bars that did not list peanut found none detectable in the 32 North American products, and detectable peanut protein in 30.8% of western European and 62% of eastern European products that carried no precautionary label, at up to 245 parts per million. (source 7)
  • Chocolate is a difficult sample. In a model dark chocolate, tempering reduced how well antibodies bound the allergens whatever extraction was used, and the choice of extraction buffer changed how much peanut protein was recovered. (source 8)

What a result does not show

That a product is free of an allergen. A negative result means none was detected, by that kit, in that sample, and processing can hide protein from the test.

The present state of the market. The survey is from 2003.

Lead isotope analysis

Also: isotopic source tracing

The proportions of lead's different isotopes, which vary with the ore the lead came from, so that lead in a food can be matched to possible sources.

In use

  • Cocoa beans from six Nigerian farms averaged 0.5 nanograms of lead per gram or less, among the lowest values reported for a natural food, while manufactured cocoa and chocolate products reached 230 and 70. The shells adsorb lead strongly, and the isotope evidence pointed to contamination after harvest, most of it during shipping or processing. (source 9)

What a result does not show

The lead content of chocolate on sale now. The samples predate 2005.

Anything about cadmium, which reaches the bean by a different route, through the soil and the tree.

In everyday use, and not covered here

These are routine in cocoa and chocolate testing. They have no entry because no source for them was read and verified for this page, and an entry written from general knowledge would look exactly like one that was checked.

  • Water activity and moisture content
  • Bean count and bean size
  • X-ray diffraction of cocoa butter crystal forms
  • Temper meters and cooling curves
  • Routine microbiology, including Salmonella testing
  • Colour and gloss measurement
  • Trace-element profiling for origin
  • Cadmium and other heavy-metal analysis

Three things the methods have in common

  • Only a tasting panel measures flavour. Everything else measures something expected to go with it, and in one trial here two lots that no bean test could separate tasted different.
  • Fast methods are predictions. Infrared and imaging methods do not measure a compound; they estimate it from a model built on other samples, and are only as good as that model’s resemblance to the sample in hand.
  • Telling groups apart is not identifying a sample. Origin and authenticity methods are shown to work by separating known sets. Assigning one unknown bar to a place, a variety or a farm is a much harder claim, and none of the studies read here makes it.

Related: the science of making chocolate, fermentation studies, the tasting vocabulary, what legally counts as chocolate and the ten genetic groups of cacao. All data.

Sources

  1. Krähmer A, Engel A, Kadow D, et al. Fast and neat: determination of biochemical quality parameters in cocoa using near infrared spectroscopy. Food Chem. 2015. doi.org/10.1016/j.foodchem.2015.02.084
  2. Caporaso N, Whitworth MB, Fowler MS, Fisk ID. Hyperspectral imaging for non-destructive prediction of fermentation index, polyphenol content and antioxidant activity in single cocoa beans. Food Chem. 2018. doi.org/10.1016/j.foodchem.2018.03.039
  3. Romero-Cortes T, Salgado-Cervantes MA, García-Alamilla P, et al. Relationship between fermentation index and other biochemical changes evaluated during the fermentation of Mexican cocoa (Theobroma cacao) beans. J Sci Food Agric. 2013. doi.org/10.1002/jsfa.6088
  4. Buchgraber M, Ulberth F, Anklam E. Cluster analysis for the systematic grouping of genuine cocoa butter and cocoa butter equivalent samples based on triglyceride patterns. J Agric Food Chem. 2004. doi.org/10.1021/jf035153v
  5. Diomande D, Antheaume I, Leroux M, et al. Multi-element, multi-compound isotope profiling as a means to distinguish the geographical and varietal origin of fermented cocoa (Theobroma cacao L.) beans. Food Chem. 2015. doi.org/10.1016/j.foodchem.2015.05.040
  6. D'Souza RN, Grimbs S, Behrends B, et al. Origin-based polyphenolic fingerprinting of Theobroma cacao in unfermented and fermented beans. Food Res Int. 2017. doi.org/10.1016/j.foodres.2017.06.007
  7. Vadas P, Perelman B. Presence of undeclared peanut protein in chocolate bars imported from Europe. J Food Prot. 2003. doi.org/10.4315/0362-028x-66.10.1932
  8. Khuda SE, Jackson LS, Fu T-J, Williams KM. Effects of processing on the recovery of food allergens from a model dark chocolate matrix. Food Chem. 2015. doi.org/10.1016/j.foodchem.2014.07.084
  9. Rankin CW, Nriagu JO, Aggarwal JK, et al. Lead contamination in cocoa and cocoa products: isotopic evidence of global contamination. Environ Health Perspect. 2005. doi.org/10.1289/ehp.8009
  10. Perotti P, Cordero C, Bortolini C, et al. Cocoa smoky off-flavor: chemical characterization and objective evaluation for quality control. Food Chem. 2020. doi.org/10.1016/j.foodchem.2019.125561
  11. Frauendorfer F, Schieberle P. Identification of the key aroma compounds in cocoa powder based on molecular sensory correlations. J Agric Food Chem. 2006. doi.org/10.1021/jf060728k
  12. Afoakwa EO, Paterson A, Fowler M, Vieira J. Influence of tempering and fat crystallization behaviours on microstructural and melting properties in dark chocolate systems. Food Res Int. 2009. doi.org/10.1016/j.foodres.2008.10.007
  13. Rodrigues JF, De Souza VR, Lima RR, et al. Temporal dominance of sensations (TDS) panel behavior: a preliminary study with chocolate. Food Qual Prefer. 2016. doi.org/10.1016/j.foodqual.2016.07.002
  14. Streule S, Freimüller Leischtfeld S, Chatelain K, et al. Effect of pod storage and drying temperature on fermentation dynamics and final bean quality of cacao Nacional in Ecuador. Foods. 2024. doi.org/10.3390/foods13101536
  15. Krause F, Steinhaus M. Off-flavor compounds in fermented cocoa: impact of overfermentation. J Agric Food Chem. 2026. doi.org/10.1021/acs.jafc.6c05685
  16. Jaimez RE, Barragan L, Fernández-Niño M, et al. Theobroma cacao L. cultivar CCN 51: a comprehensive review. PeerJ. 2022. (Read in full.) doi.org/10.7717/peerj.12676

Reviewed 8 October 2026.