{
  "id": "roast-quality-control",
  "url": "https://coffearium.org/entries/roast-quality-control/",
  "license": "CC-BY-4.0",
  "title": "Roast quality control",
  "summary": "How roasteries check that each batch matches its target – roast colour, roast loss, moisture, appearance and flavour against limits and a reference – with food-safety checks and records before the coffee is released.",
  "domain": "roasting",
  "kind": "Method",
  "stage": "roast",
  "topics": [],
  "topicSections": {},
  "lang": "en",
  "status": "verified",
  "reviewer": "Coffearium verification pass",
  "lastVerified": "2026-11-16T00:00:00.000Z",
  "literatureSearched": "2026-11-16T00:00:00.000Z",
  "updated": "2026-11-16T00:00:00.000Z",
  "added": "2026-11-16T00:00:00.000Z",
  "brief": [
    "Colour is checked with roast meters on ground coffee, with grind and temperature controlled; meter scales are arbitrary, and brands do not fully agree.<sup><a href=\"#src-s1\">1</a>,<a href=\"#src-s2\">2</a></sup>",
    "Roasted coffee usually holds 1–2.5% moisture (basis not stated). Some national food rules set a maximum; Switzerland and Austria, for example, allow 5% by mass.<sup><a href=\"#src-s3\">3</a>,<a href=\"#src-s7\">7</a>,<a href=\"#src-s8\">8</a></sup>",
    "The triangle test (ISO 4120) shows whether a change in ingredients or processing gives a perceptible sensory difference.<sup><a href=\"#src-s14\">14</a></sup>",
    "EU roasters must run their own acrylamide sampling and analysis; the benchmark level for roast coffee is 400 µg/kg.<sup><a href=\"#src-s18\">18</a></sup>",
    "In a pilot-roaster study, near-infrared spectra with multivariate statistical process control detected all imposed disturbances during the roast.<sup><a href=\"#src-s10\">10</a></sup>"
  ],
  "quickFacts": [
    {
      "label": "Main roast-degree check",
      "value": "Colour of ground coffee with a roast meter; grind and temperature controlled"
    },
    {
      "label": "Other indices",
      "value": "Roast loss, whole-bean colour, bulk density, dry-matter loss"
    },
    {
      "label": "Moisture, reference method",
      "value": "ISO 11294, loss in mass at 103 °C, roasted ground coffee"
    },
    {
      "label": "Water, maximum in national rules (examples)",
      "value": "5% by mass (Swiss beverage ordinance; Austrian Food Codex)"
    },
    {
      "label": "Sensory checks",
      "value": "Cupping; triangle and duo-trio tests against regular production"
    },
    {
      "label": "EU acrylamide benchmark, roast coffee",
      "value": "400 µg/kg (Regulation (EU) 2017/2158)"
    }
  ],
  "aka": [
    "Roasting quality control",
    "Roast QC",
    "Quality control in roasting"
  ],
  "linkTerms": [
    "roast quality control",
    "roasting quality control",
    "roast QC",
    "production cupping",
    "batch release"
  ],
  "related": [
    "roast-degree-and-colour",
    "roast-loss",
    "roast-defects",
    "roast-profile",
    "volume-and-bulk-density",
    "cooling-and-quenching",
    "cupping",
    "sensory-science-methods",
    "roasting-contaminants",
    "green-coffee-quality-control",
    "sample-roasters",
    "traceability",
    "blending-before-or-after-roasting"
  ],
  "contested": [],
  "corrections": [
    {
      "date": "2026-11-16T00:00:00.000Z",
      "change": "SCA 102-2024 is now described as defining cupping roast by ground-coffee lightness and by a reading of 63 on the Agtron Gourmet scale or its equivalent on other meters, instead of allowing infrared meters with their own targets.",
      "reason": "The standard gives one roast-meter target with equivalents; meter-specific targets come from the CVA document."
    },
    {
      "date": "2026-11-16T00:00:00.000Z",
      "change": "Swiss and Austrian limits are presented as national examples in the brief, quick facts, text, table and caption; the quick fact on charred beans was removed; the Austrian Food Codex is no longer called a legal maximum.",
      "reason": "International scope; the Austrian Food Codex is an official expert opinion, not a statute."
    }
  ],
  "sources": [
    {
      "id": "s1",
      "title": "SCA Standard 102-2024 Coffee Value Assessment: Sample Preparation and Tasting Mechanics",
      "authors": "Specialty Coffee Association",
      "year": 2024,
      "publisher": "Specialty Coffee Association",
      "url": "https://sca.coffee/s/AW_SCA-102_Sample-Preparation_281024_Secured.pdf",
      "tier": "A/B",
      "tierNote": "A for what the standard specifies; B for its explanations",
      "quote": "When using roast meters, parameters such as the grind level and the coffee temperature should be controlled.",
      "quotes": [
        {
          "text": "For the purposes of this standard, cupping level roast shall be defined in colorimetry terms as lightness of the ground coffee (L* in the CIELAB coordinates) of 26–29 (corresponding approximately to Agtron/SCA Roast Color Classification System tile #65).",
          "for": "cupping roast as a target range"
        },
        {
          "text": "Roast batches that do not fall in the desired roast level range or that present underdeveloped, burnt, or baked characteristics should not be used for cupping, and the roast batch should be repeated.",
          "for": "batches outside the range are repeated"
        },
        {
          "text": "It should be considered that samples with a lot of chaff will show a larger error.",
          "for": "chaff and meter readings"
        },
        {
          "text": "In roast-meter terms, cupping level roast shall be defined as a target reading of 63 in the Agtron \"Gourmet\" scale or its equivalent in other roast meters.",
          "for": "roast-meter target (63, Agtron Gourmet scale)"
        },
        {
          "text": "Alternative methods for roast level control, such as dry matter loss, or volume increase of the coffee bean, may be used if correlations with colorimetry or spectrometry have been established.",
          "for": "alternatives to colour"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "SCA standard for preparing cupping samples in the Coffee Value Assessment, 2024; section 5.1 Roasting. It governs cupping roasts, not production batches.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s2",
      "title": "A System to Assess Coffee Value: Understanding the Specialty Coffee Association’s Coffee Value Assessment (June 2024)",
      "authors": "Specialty Coffee Association (writing team Fernández-Alduenda, M.; Ionescu, K. E.; Giuliano, P.; Rugolo, J.; Delrue, R.)",
      "year": 2024,
      "publisher": "Specialty Coffee Association",
      "url": "https://static1.squarespace.com/static/584f6bbef5e23149e5522201/t/667182ffdde8a5081afc2d8c/1718715138872/SCA+-+A+System+to+Asssess+Coffee+Value+-+June+2024+(Secured).pdf",
      "tier": "A/B",
      "tierNote": "A for what the CVA itself specifies (procedure, roast-meter targets in Table 4); B for its explanations",
      "quote": "However, the scales used by the different roast meters are arbitrary, and even if the same scale is used, sometimes different brands will not completely correlate.",
      "quotes": [
        {
          "text": "Table 4 lists the target reading for cupping roast level using different brands of roast meters.",
          "for": "meter-specific targets"
        },
        {
          "text": "Unfortunately, to the best of our knowledge, there are no simple measurements to control for these problems across different roasting technologies and coffee types.",
          "for": "no simple test for too-fast or too-slow roasts"
        },
        {
          "text": "However, cuppers may be trained to detect these problems by assessing the beans’ structure, gradient of colors, and flavor.",
          "for": "detection by cuppers"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "SCA explanatory document on the Coffee Value Assessment, June 2024; section 5.1.1 on sample roasting, roast meters (Table 4) and roasting problems.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s3",
      "title": "Roasting Conditions and Coffee Flavor: A Multi-Study Empirical Investigation",
      "authors": "Münchow, M.; Alstrup, J.; Steen, I.; Giacalone, D.",
      "year": 2020,
      "publisher": "Beverages 6(2), 29",
      "url": "https://www.mdpi.com/2306-5710/6/2/29",
      "doi": "10.3390/beverages6020029",
      "tier": "A",
      "quote": "This could lead to commercial failures as, for instance, colour measurements are often left out of QC procedures in specialty coffee roasteries; if colour is indeed an important flavour modulator, this could result in consumer relevant, but unknown (not checked) batch-to-batch variations.",
      "quotes": [
        {
          "text": "During the roasting process, the beans lose between 15 and 22% of their weight, of which most is water as it ends up as a rather dry product with between 1 and 2.5% moisture",
          "for": "roast loss and moisture of roasted coffee"
        },
        {
          "text": "The results indicated that, while both parameters significantly affected coffee flavour, colour was the stronger predictor of the two.",
          "for": "colour predicts flavour better than time"
        }
      ],
      "scope": {
        "claim": "description",
        "context": "Paper combining eight sensory studies; only Studies 1 and 2 varied both colour and time (Denmark, 2020). The QC remark is the authors' observation, not a survey result; roast loss and moisture are cited from earlier work.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s4",
      "title": "Use of colour parameters for roasted coffee assessment",
      "authors": "Bicho, N. C.; Leitão, A. E.; Ramalho, J. C.; Lidon, F. C.",
      "year": 2012,
      "publisher": "Ciência e Tecnologia de Alimentos 32(3), 436–442",
      "url": "https://www.redalyc.org/pdf/3959/395940113004.pdf",
      "doi": "10.1590/S0101-20612012005000068",
      "tier": "A",
      "quote": "They showed higher values of brightness and chromatic coordinates, mainly in Robusta coffee, yet the colour difference decreased with the roasting intensity, possibly as a result of the temperature gradient across the bean, which inevitably leads to colour deviation in relation to ground roasted coffee.",
      "quotes": [
        {
          "text": "During the roasting process, weight loss usually varies between 14-23% depending on the botanical origin, green coffee moisture, storage conditions, and the roasting method.",
          "for": "roast loss range"
        }
      ],
      "scope": {
        "claim": "finding",
        "context": "Laboratory roasts of one arabica and one robusta at three intensities (200–240 °C, 5–12 min); colour of whole and ground beans with a Minolta CR-300 colorimeter, illuminants D65 and C (Portugal, 2012).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s5",
      "title": "ISO 11294:1994 Roasted ground coffee — Determination of moisture content — Method by determination of loss in mass at 103 °C (Routine method)",
      "authors": "International Organization for Standardization (ISO/TC 34/SC 15)",
      "year": 1994,
      "publisher": "ISO",
      "url": "https://www.iso.org/standard/19272.html",
      "tier": "A",
      "tierNote": "Standard; only the abstract and status on the ISO catalogue page were read (last reviewed and confirmed in 2018, current as of November 2026). The full text was not consulted.",
      "quote": "Specifies a routine method for the determintion of loss in mass at 103 °C of roasted ground coffee.",
      "quotes": [
        {
          "text": "Most suited to degassed roasted ground coffee, because of the presence of volatile matter, especially carbon dioxide, in variable quantities in roasted coffee.",
          "for": "degassed samples; carbon dioxide"
        }
      ],
      "scope": {
        "claim": "definition",
        "context": "Scope of the ISO routine method for moisture in roasted ground coffee (1994, confirmed 2018).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s6",
      "title": "ISO 6669:1995 Green and roasted coffee — Determination of free-flow bulk density of whole beans (Routine method)",
      "authors": "International Organization for Standardization (ISO/TC 34/SC 15)",
      "year": 1995,
      "publisher": "ISO (preview pages via ANSI Webstore)",
      "url": "https://webstore.ansi.org/preview-pages/ISO/preview_ISO+6669-1995.pdf",
      "tier": "A",
      "tierNote": "Standard; A for what it specifies and defines. Only the cover, foreword and introduction were openly available (same source as in Bean size and density).",
      "quote": "Bulk density is defined as the ratio of mass to volume occupied.",
      "quotes": [
        {
          "text": "Botanic, horticultural, processing, storage and handling factors, including the effect of age, variously influence the bulk density of green coffee beans, whilst their roasting behaviour and conditions additionally influence the bulk density of the roasted beans.",
          "for": "roasting conditions influence bulk density"
        },
        {
          "text": "Knowledge of the bulk density of green and of roasted whole coffee beans is important for their trade, since it determines the volume occupied by a given mass of beans, which is a factor in their packaging, storage and transport.",
          "for": "why bulk density is measured"
        }
      ],
      "scope": {
        "claim": "definition",
        "context": "Introduction of the ISO routine method for free-flow bulk density of whole green and roasted beans (1995).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s7",
      "title": "Verordnung des EDI über Getränke vom 16. Dezember 2016 (SR 817.022.12), Art. 43",
      "authors": "Eidgenössisches Departement des Innern (EDI)",
      "year": 2016,
      "publisher": "Swiss Confederation (consolidated version of 1 July 2020, via LexFind)",
      "url": "https://www.lexfind.ch/tolv/211066/de",
      "tier": "A",
      "tierNote": "Consolidated version of 1 July 2020 as shown on LexFind. Fedlex could not be read here (needs JavaScript). The only later amendment found, explained by the FSVO on 28 May 2025, changes wine articles (Art. 61, 75, 76, 77–79, 161b, Annex 9), not Art. 42–43; checked 16 Nov 2026.",
      "quote": "Röstkaffee darf nicht mehr als 1 Massenprozent verkohlte Bohnen und nicht mehr als 5 Massenprozent Wasser enthalten.",
      "quotes": [],
      "scope": {
        "claim": "guidance",
        "context": "Swiss ordinance on beverages, Art. 43 (requirements for roasted coffee); legal limits for charred beans and water in roasted coffee sold in Switzerland.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s8",
      "title": "Österreichisches Lebensmittelbuch, IV. Auflage, Codexkapitel B 12 Kaffee, Kaffeemittel",
      "authors": "Bundesministerium für Soziales, Gesundheit, Pflege und Konsumentenschutz (Codexkommission)",
      "year": 2015,
      "publisher": "Republic of Austria (version with amendments of 2017 and 2020, via WKO)",
      "url": "https://news.wko.at/oe/gewerbe-handwerk/lebensmittelgewerbe/codex-b12-kaffee-aenderungen.pdf",
      "tier": "A",
      "tierNote": "Official Austrian Food Codex (published by the ministry; an objectified expert opinion, not a statute). Chapter B 12 as published 18 March 2015 with amendments of 22 Dec 2017 and 20 July 2020; the WKO list of Codex updates (30 Sep 2026) shows no later change to B 12; checked 16 Nov 2026.",
      "quote": "Sein Wassergehalt darf 5 Gew.-%, sein Aschegehalt 6,5 Gew.-% nicht übersteigen",
      "quotes": [
        {
          "text": "Richtig geröstete Kaffeebohnen zeigen eine gleichmäßig braune Farbe und sind auch im Inneren weder licht noch verbrannt, sondern braun.",
          "for": "evenly roasted beans (1.2.1.1)"
        },
        {
          "text": "Ehe gerösteter Kaffee in den Handel kommt, wird er meistens verlesen, wobei sowohl missfarbige Bohnen als auch Fremdkörper („Einwurf“) soweit wie möglich entfernt werden.",
          "for": "sorting out discoloured beans and foreign bodies (1.2.1.3)"
        },
        {
          "text": "Der Einwurf bei in Verkehr gebrachtem Röstkaffee darf 2 Gew.-% nicht übersteigen.",
          "for": "Einwurf at most 2% (1.2.1.3)"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "Austrian Food Codex chapter on coffee, section 1.2.1 (roasted coffee): description of a correct roast, sorting, limits for water and for discoloured beans and foreign bodies.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s9",
      "title": "NIST/SEMATECH e-Handbook of Statistical Methods, section 6.3.1: What are Control Charts?",
      "authors": "National Institute of Standards and Technology (NIST); SEMATECH",
      "publisher": "U.S. Department of Commerce, NIST",
      "url": "https://www.itl.nist.gov/div898/handbook/pmc/section3/pmc31.htm",
      "tier": "A",
      "quote": "In general, the chart contains a center line that represents the mean value for the in-control process.",
      "quotes": [
        {
          "text": "Two other horizontal lines, called the upper control limit (UCL) and the lower control limit (LCL), are also shown on the chart.",
          "for": "upper and lower control limits"
        },
        {
          "text": "These control limits are chosen so that almost all of the data points will fall within these limits as long as the process remains in-control.",
          "for": "purpose of the limits"
        },
        {
          "text": "Usually this multiple is 3 and thus the limits are called 3-sigma limits.",
          "for": "3-sigma limits"
        }
      ],
      "scope": {
        "claim": "definition",
        "context": "General statistics handbook of a US government institute; definition of univariate control charts, not specific to coffee.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s10",
      "title": "Real-time monitoring of a coffee roasting process with near infrared spectroscopy using multivariate statistical analysis: A feasibility study",
      "authors": "Catelani, T. A.; Santos, J. R.; Páscoa, R. N. M. J.; Pezza, L.; Pezza, H. R.; Lopes, J. A.",
      "year": 2018,
      "publisher": "Talanta 179, 292–299",
      "url": "https://repositorio.unesp.br/bitstream/11449/179372/1/2-s2.0-85034610548.pdf",
      "doi": "10.1016/j.talanta.2017.11.010",
      "tier": "A",
      "quote": "A PCA model encompassing a time window of four minutes with three principal components was able to efficiently detect all disturbances assayed.",
      "quotes": [
        {
          "text": "This methodology was tested on a set of five batches where disturbances of different nature were imposed to simulate real faulty situations.",
          "for": "five disturbed batches"
        },
        {
          "text": "A total of fifteen roasting batches were defined according to an experimental design to develop the MSPC models.",
          "for": "fifteen batches for the model"
        },
        {
          "text": "However, this control is not very effective because process disturbances are not or only detected at the end.",
          "for": "conventional control detects disturbances late"
        },
        {
          "text": "NIR spectroscopy combined with the MSPC approach proved to be an adequate auxiliary tool for coffee roasters to detect faults in a conventional roasting process in real-time.",
          "for": "conclusion"
        }
      ],
      "scope": {
        "claim": "finding",
        "context": "Pilot-scale drum roaster (Hottop KN8828B-2K), batches of about 250 g, blends of Brazilian and Timorese arabica and Ugandan robusta; one NIR spectrum per minute; Portugal and Brazil, 2018.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s11",
      "title": "In-line monitoring of the coffee roasting process with near infrared spectroscopy: Measurement of sucrose and colour",
      "authors": "Santos, J. R.; Viegas, O.; Páscoa, R. N. M. J.; Ferreira, I. M. P. L. V. O.; Rangel, A. O. S. S.; Lopes, J. A.",
      "year": 2016,
      "publisher": "Food Chemistry 208, 103–110",
      "url": "https://repositorio.ucp.pt/entities/publication/a83ce3fd-c0a9-4d02-a459-fd3ddd63f0e1/full",
      "doi": "10.1016/j.foodchem.2016.03.114",
      "tier": "A",
      "quote": "All near infrared spectroscopy-based calibrations were developed resorting to partial least squares regression.",
      "quotes": [
        {
          "text": "coefficient of determination higher than 10 and 0.85 respectively, for all modelled parameters.",
          "for": "RER above 10 and R² above 0.85 (abstract; the preceding words are “range-error-ratio and”)"
        }
      ],
      "abstractOnly": true,
      "scope": {
        "claim": "finding",
        "context": "Abstract only (read on the Universidade Católica Portuguesa repository page): in-line NIR during roasting of arabica and robusta from Brazil, East Timor, India and Uganda, slow and fast profiles; RER above 10 and R² above 0.85 for all modelled parameters (Portugal, 2016).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s12",
      "title": "Prediction of coffee aroma from single roasted coffee beans by hyperspectral imaging",
      "authors": "Caporaso, N.; Whitworth, M. B.; Fisk, I. D.",
      "year": 2022,
      "publisher": "Food Chemistry 371, 131159",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8617352",
      "doi": "10.1016/j.foodchem.2021.131159",
      "tier": "A",
      "quote": "However, coffee batches are not homogeneous and, in many cases, contain both within specification coffee beans and beans with unique flavour profiles.",
      "quotes": [
        {
          "text": "This study applied hyperspectral imaging (1000–2500 nm) to predict volatile compounds in single roasted coffee beans, as measured by Solid Phase Micro Extraction-Gas Chromatography-Mass Spectrometry and Gas Chromatography-Olfactometry.",
          "for": "single-bean hyperspectral imaging"
        },
        {
          "text": "Ultimately, the application of this insight could enable better standardisation of roast coffee quality, removal of defect beans and even selection of the most distinctive batches of coffee beans for special target markets.",
          "for": "possible quality-control uses"
        }
      ],
      "scope": {
        "claim": "hypothesis",
        "context": "Laboratory study, single beans roasted individually (210 °C, 3 min), United Kingdom, 2022; the QC uses are the authors' outlook, not tested in production.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s13",
      "title": "Using hyperspectral imaging to characterize consistency of coffee brands and their respective roasting classes",
      "authors": "Nansen, C.; Singh, K.; Mian, A.; Allison, B. J.; Simmons, C. W.",
      "year": 2016,
      "publisher": "Journal of Food Engineering 190, 34–39",
      "url": "https://agris.fao.org/search/es/records/65de2acb63b8185d9ca74b60",
      "doi": "10.1016/j.jfoodeng.2016.06.010",
      "tier": "A",
      "quote": "In this study, we characterized the consistency (levels of extractable protein and reducing sugars) of 15 brands of roasted coffee beans, which were obtained from a supermarket at two dates about six months apart.",
      "quotes": [
        {
          "text": "The uniqueness and consistency of commercial food and beverage brands are critically important for their marketability.",
          "for": "consistency of brands"
        },
        {
          "text": "Coffee brands varied markedly in extractable protein and reducing sugar contents between dates.",
          "for": "chemistry differed markedly between dates"
        }
      ],
      "abstractOnly": true,
      "scope": {
        "claim": "finding",
        "context": "Abstract only (read on AGRIS): 15 supermarket roasted-coffee brands bought twice about six months apart (2016); brands varied markedly between dates in extractable protein and reducing sugars.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s14",
      "title": "ISO 4120:2021 Sensory analysis — Methodology — Triangle test",
      "authors": "International Organization for Standardization",
      "year": 2021,
      "publisher": "ISO",
      "url": "https://www.iso.org/standard/76666.html",
      "tier": "A",
      "tierNote": "Only the abstract (scope) on the ISO catalogue page was read; last reviewed and confirmed in 2026. The full text and its tables were not consulted.",
      "quote": "The method is a forced-choice procedure.",
      "quotes": [
        {
          "text": "when, for example, a change is made in ingredients, processing, packaging, handling or storage;",
          "for": "uses of the triangle test"
        },
        {
          "text": "The method is applicable only if the products are homogeneous.",
          "for": "homogeneous products"
        },
        {
          "text": "The method is statistically more efficient than the duo-trio test (described in ISO 10399), but has limited use with products that exhibit strong carryover and/or lingering flavours.",
          "for": "carryover limits"
        }
      ],
      "scope": {
        "claim": "definition",
        "context": "Scope of the ISO triangle test, 2021 edition (confirmed 2026).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s15",
      "title": "ISO 10399:2017 Sensory analysis — Methodology — Duo-trio test",
      "authors": "International Organization for Standardization",
      "year": 2017,
      "publisher": "ISO",
      "url": "https://committee.iso.org/standard/74219.html",
      "tier": "A",
      "tierNote": "Only the openly visible abstract and status on the ISO catalogue page were read. The page lists the 2017 edition as withdrawn on 15 May 2026 and replaced by ISO 10399:2026 (edition 4, published 15 May 2026), as of November 2026. The 2026 abstract (committee.iso.org/standard/88370.html) contains both quoted passages word for word.",
      "quote": "The method is statistically less efficient than the triangle test (described in ISO 4120) but is easier to perform by the assessors.",
      "quotes": [
        {
          "text": "the constant-reference technique, used when one product is familiar to the assessors (e.g. a sample from regular production);",
          "for": "constant reference from regular production"
        }
      ],
      "scope": {
        "claim": "definition",
        "context": "Scope of the ISO duo-trio test, 2017 edition (withdrawn May 2026, replaced by ISO 10399:2026).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s16",
      "title": "Coffee Flavor: A Review",
      "authors": "Seninde, D. R.; Chambers, E., IV",
      "year": 2020,
      "publisher": "Beverages 6(3), 44",
      "url": "https://www.mdpi.com/2306-5710/6/3/44",
      "doi": "10.3390/beverages6030044",
      "tier": "A",
      "quote": "Cupping is differentiated from descriptive sensory analysis by its use of industry experts who have coffee quality control experience and judge coffee quality on a quality scale.",
      "quotes": [],
      "scope": {
        "claim": "description",
        "context": "Review of coffee flavour and its sensory assessment (United States, 2020).",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s17",
      "title": "Regulation (EC) No 852/2004 of the European Parliament and of the Council of 29 April 2004 on the hygiene of foodstuffs",
      "authors": "European Parliament and Council of the European Union",
      "year": 2004,
      "publisher": "EUR-Lex (original text as published)",
      "url": "https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32004R0852",
      "tier": "A",
      "tierNote": "Original text. Article 5(1)–(2). Later amendments were not checked for these articles.",
      "quote": "identifying any hazards that must be prevented, eliminated or reduced to acceptable levels;",
      "quotes": [
        {
          "text": "Food business operators shall put in place, implement and maintain a permanent procedure or procedures",
          "for": "HACCP-based procedures (Art. 5(1))"
        },
        {
          "text": "establishing critical limits at critical control points which separate acceptability from unacceptability",
          "for": "critical limits (Art. 5(2)(c))"
        },
        {
          "text": "establishing corrective actions when monitoring indicates that a critical control point is not under control;",
          "for": "corrective actions (Art. 5(2)(e))"
        },
        {
          "text": "establishing documents and records commensurate with the nature and size of the food business",
          "for": "records (Art. 5(2)(g))"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "EU hygiene regulation, Article 5: HACCP-based procedures for food business operators after primary production.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s18",
      "title": "Commission Regulation (EU) 2017/2158 of 20 November 2017 establishing mitigation measures and benchmark levels for the reduction of the presence of acrylamide in food",
      "authors": "European Commission",
      "year": 2017,
      "publisher": "Official Journal of the European Union L 304, 21 November 2017 (EUR-Lex)",
      "url": "https://eur-lex.europa.eu/eli/reg/2017/2158/oj/eng",
      "tier": "A",
      "tierNote": "Regulation; cited for what it requires and sets. The EUR-Lex pages did not render the text here (the consolidated text is blocked to automated reading); recital 10, Articles 1, 4 and 6, Annex I (coffee), Annex III and Annex IV were read in the legislation.gov.uk copy of the original as adopted. EUR-Lex lists no amending act; a Commission review of the acrylamide framework (new benchmark and maximum levels) was in consultation in 2026 but not adopted as of 16 Nov 2026.",
      "quote": "FBOs shall identify the critical roast conditions to ensure minimal acrylamide formation within the target flavour profile.",
      "quotes": [
        {
          "text": "Food business operators referred to in Article 2(1) shall establish a programme for their own sampling and analysis of the levels of acrylamide in the foodstuffs listed in Article 1(2).",
          "for": "own sampling programme (Art. 4(1))"
        },
        {
          "text": "If the sampling and analysis results indicate that the levels are not below the benchmark levels of acrylamide set out in Annex IV, food business operators referred to in Article 2(1) and (3) shall review without delay the mitigation measures in accordance with Article 2(4).",
          "for": "review when not below the benchmark (Art. 4(5))"
        },
        {
          "text": "FBOs shall ensure that a representative sample of each product type is taken for analysis of acrylamide concentration.",
          "for": "representative sample of each product type (Annex III)"
        },
        {
          "text": "The sample shall be representative for the sampled batch.",
          "for": "representative of the batch (Annex III)"
        },
        {
          "text": "For benchmark level ≥ 125 μg/kg: ≤ 50 μg/kg",
          "for": "limit of quantification (Annex III)"
        },
        {
          "text": "75-110 %",
          "for": "recovery (Annex III, table cell)"
        },
        {
          "text": "Three tenths of LOQ",
          "for": "limit of detection (Annex III, table cell)"
        },
        {
          "text": "Control of roast conditions shall be incorporated into a Pre-requisite Program (PRP) as part of Good Manufacturing Practice (GMP).",
          "for": "prerequisite programme (Annex I)"
        },
        {
          "text": "Benchmark levels are performance indicators to be used to verify the effectiveness of the mitigation measures",
          "for": "benchmark levels as performance indicators (recital 10)"
        },
        {
          "text": "FBOs shall, undertake sampling and analysis at least annually for products that have a known and well-controlled acrylamide level",
          "for": "at least annual sampling (Annex III; opening of a longer sentence)"
        },
        {
          "text": "products based on Robusta beans tend to have higher acrylamide levels than products based on Arabica beans",
          "for": "robusta in blends (Annex I)"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "EU regulation on acrylamide mitigation, in force since 11 April 2018; obligations of food business operators that roast coffee; benchmark levels are performance indicators, not maximum levels.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s19",
      "title": "Regulation (EC) No 178/2002 of the European Parliament and of the Council of 28 January 2002 laying down the general principles and requirements of food law, establishing the European Food Safety Authority and laying down procedures in matters of food safety",
      "authors": "European Parliament and Council of the European Union",
      "year": 2002,
      "publisher": "Official Journal of the European Communities, L 31 (EUR-Lex)",
      "url": "https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32002R0178",
      "tier": "A",
      "tierNote": "Original text as published; Article 18 (traceability). Later amendments were not checked for this article.",
      "quote": "Food and feed business operators shall be able to identify any person from whom they have been supplied with a food",
      "quotes": [
        {
          "text": "Food and feed business operators shall have in place systems and procedures to identify the other businesses",
          "for": "one step forward (Art. 18(3))"
        }
      ],
      "scope": {
        "claim": "guidance",
        "context": "EU General Food Law, Article 18: traceability one step back and one step forward for all food businesses.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s20",
      "title": "AP-42, Fifth Edition, Section 9.13.2: Coffee Roasting",
      "authors": "U.S. Environmental Protection Agency",
      "year": 1995,
      "publisher": "U.S. EPA, Compilation of Air Pollutant Emission Factors",
      "url": "https://www.epa.gov/sites/default/files/2020-10/documents/c9s13-2.pdf",
      "tier": "A",
      "quote": "Destoners are air classifiers that remove stones, metal fragments, and other waste not removed during initial screening from the beans.",
      "quotes": [
        {
          "text": "Bags of green coffee beans are hand- or machine-opened, dumped into a hopper, and screened to remove debris.",
          "for": "screening of green coffee"
        }
      ],
      "scope": {
        "claim": "description",
        "context": "US EPA description of the coffee roasting process for emission factors (1995); describes equipment, not quality limits.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    },
    {
      "id": "s21",
      "title": "Triangle Test",
      "authors": "Society of Sensory Professionals",
      "publisher": "Society of Sensory Professionals, Sensory Wiki",
      "url": "https://www.sensorysociety.org/knowledge/sspwiki/Pages/Triangle%20Test.aspx",
      "tier": "B",
      "quote": "During a triangle test, a panelist is presented with one different and two alike samples.",
      "quotes": [
        {
          "text": "The panelist is instructed to identify the odd sample and record his answer.",
          "for": "pick the odd sample"
        }
      ],
      "scope": {
        "claim": "description",
        "context": "Wiki page of a professional society for sensory science; describes the triangle-test procedure in general terms.",
        "checked": "2026-11-16T00:00:00.000Z"
      },
      "accessed": "2026-11-16T00:00:00.000Z"
    }
  ],
  "markdown": "## What roast quality control does\n\nA roastery sells the same coffee again and again. Buyers expect each bag to look and taste like the last one. Roast quality control (QC) checks each batch – one load of the roaster – against a target before the coffee leaves the plant.\n\nThe target is set for each product. It usually names a roast degree, measured as colour, and a flavour, kept as a reference sample. The Specialty Coffee Association (SCA) gives one example for cupping roasts: a colour range of L\\* 26–29 on ground coffee.<sup><a href=\"#src-s1\">1</a></sup> Batches outside the range, or with burnt or baked character, are not used, and the roast is repeated.<sup><a href=\"#src-s1\">1</a></sup>\n\nVariation is real. In a 2016 study, 15 brands of roasted coffee were bought from a supermarket twice, about six months apart. Their chemistry differed markedly between the two dates.<sup><a href=\"#src-s13\">13</a></sup> A 2020 sensory paper notes that specialty roasteries often leave colour out of QC. Batch-to-batch differences can then go unnoticed.<sup><a href=\"#src-s3\">3</a></sup>\n\nQC also covers food safety. In the European Union (EU), roasteries run food-safety procedures based on hazard analysis.<sup><a href=\"#src-s17\">17</a></sup> They also sample and analyse their coffee for acrylamide, a contaminant formed in roasting.<sup><a href=\"#src-s18\">18</a></sup>\n\n## Checks along the line\n\nChecks happen at several points between green coffee and dispatch. The figure follows one batch. Green coffee is checked on arrival (see Green coffee quality control); this entry starts at the roaster.\n\n<figure class=\"iso-fig iso-fig--below\">\n<div data-figure=\"roast-quality-control-checkpoints\"></div>\n<ol class=\"fig-legend\">\n<li data-n=\"1\"><span><b>Cleaning.</b> Green coffee is screened to remove debris before roasting.<sup><a href=\"#src-s20\">20</a></sup></span></li>\n<li data-n=\"2\"><span><b>During the roast.</b> Roast conditions are controlled; in research, near-infrared spectra follow the batch in real time.<sup><a href=\"#src-s10\">10</a>,<a href=\"#src-s11\">11</a>,<a href=\"#src-s18\">18</a></sup></span></li>\n<li data-n=\"3\"><span><b>After cooling.</b> A destoner removes stones, metal fragments and other waste.<sup><a href=\"#src-s20\">20</a></sup></span></li>\n<li data-n=\"4\"><span><b>Batch checks.</b> Colour of whole and ground coffee, roast loss, moisture and appearance against the target. A periodic sample goes for acrylamide analysis.<sup><a href=\"#src-s1\">1</a>,<a href=\"#src-s3\">3</a>,<a href=\"#src-s5\">5</a>,<a href=\"#src-s8\">8</a>,<a href=\"#src-s18\">18</a></sup></span></li>\n<li data-n=\"5\"><span><b>Cup.</b> Cupping, or a difference test against a sample from regular production.<sup><a href=\"#src-s14\">14</a>,<a href=\"#src-s15\">15</a>,<a href=\"#src-s16\">16</a></sup></span></li>\n<li data-n=\"6\"><span><b>Release or hold.</b> Records are kept; batches outside the limits are held and corrected.<sup><a href=\"#src-s17\">17</a>,<a href=\"#src-s19\">19</a></sup></span></li>\n</ol>\n<figcaption>Quality control checkpoints for one batch in a roastery, from green coffee to dispatch. Schematic, not to scale; the order of checks after cooling varies between plants.<sup><a href=\"#src-s17\">17</a>,<a href=\"#src-s20\">20</a></sup></figcaption>\n</figure>\n\n## What is measured\n\nEach check answers a different question. Together they describe roast degree, dryness, evenness and flavour.\n\n- **Colour.** Colour is the usual measure of roast degree (see Roast degree and colour). A roast meter reads it on ground coffee; the SCA asks that grind level and sample temperature be controlled.<sup><a href=\"#src-s1\">1</a></sup> Meter scales are arbitrary, and two brands may not fully agree.<sup><a href=\"#src-s2\">2</a></sup>\n- **Whole bean and ground.** Whole beans read darker than the same beans ground. In a 2012 study, the gap shrank as the roast got darker.<sup><a href=\"#src-s4\">4</a></sup> A gap that is larger than usual can point to an uneven roast through the bean (our reading).\n- **Roast loss.** This is the weight lost in roasting. Sources give 15–22% and 14–23% for typical roasts (see Roast loss).<sup><a href=\"#src-s3\">3</a>,<a href=\"#src-s4\">4</a></sup>\n- **Moisture.** Roasted coffee usually holds 1–2.5% water (basis not stated).<sup><a href=\"#src-s3\">3</a></sup> Some national food rules set a maximum: the Swiss beverage ordinance and the Austrian Food Codex, for example, allow at most 5% by mass.<sup><a href=\"#src-s7\">7</a>,<a href=\"#src-s8\">8</a></sup>\n- **Bulk density.** This is mass per volume of loose beans. Roasting conditions change it, and it matters for packaging (see Volume and bulk density).<sup><a href=\"#src-s6\">6</a></sup>\n- **Appearance.** According to the Austrian Food Codex, a correct roast is evenly brown and brown inside, neither pale nor burnt.<sup><a href=\"#src-s8\">8</a></sup> The Swiss ordinance also allows at most 1% charred beans by mass.<sup><a href=\"#src-s7\">7</a></sup>\n\n<figure>\n<div data-figure=\"roast-quality-control-limits\"></div>\n<figcaption>Moisture reported for roasted coffee, against the water maximum in two national food rules, given as examples (Switzerland, Austria). Values in % of mass.<sup><a href=\"#src-s3\">3</a>,<a href=\"#src-s7\">7</a>,<a href=\"#src-s8\">8</a></sup></figcaption>\n</figure>\n\nRoast defects explains what uneven, scorched or baked beans look like.\n\n## Tasting against a reference\n\nInstruments do not taste. Flavour is checked in the cup, usually by cupping. According to a 2020 review, cupping relies on industry experts with coffee quality-control experience.<sup><a href=\"#src-s16\">16</a></sup>\n\nDifference tests ask a narrower question: is this batch different from the usual one? In a triangle test, tasters get three samples, two of them alike, and must pick the odd one.<sup><a href=\"#src-s14\">14</a>,<a href=\"#src-s21\">21</a></sup> ISO 4120 names a change in ingredients, processing, packaging, handling or storage as a typical use.<sup><a href=\"#src-s14\">14</a></sup> In a duo-trio test, the reference can be a sample from regular production.<sup><a href=\"#src-s15\">15</a></sup> Sensory science methods describes both tests.\n\n## Food safety after roasting\n\nFood-safety checks follow a plan. Under EU hygiene law, operators identify hazards, set critical limits and act when monitoring shows a step is out of control.<sup><a href=\"#src-s17\">17</a></sup>\n\n**Foreign bodies.** Green coffee is screened for debris before roasting.<sup><a href=\"#src-s20\">20</a></sup> After roasting, a destoner – an air classifier – removes stones, metal fragments and other waste.<sup><a href=\"#src-s20\">20</a></sup>\n\n**Acrylamide.** EU roasters must identify the roast conditions that keep acrylamide low within the target flavour.<sup><a href=\"#src-s18\">18</a></sup> They run their own sampling and analysis programme.<sup><a href=\"#src-s18\">18</a></sup> The benchmark level for roast coffee is 400 µg/kg; it is a performance indicator, not a maximum level.<sup><a href=\"#src-s18\">18</a></sup> Roasting contaminants covers the chemistry.\n\n## Records and release\n\nEach batch leaves a trail. EU hygiene law asks for documents and records that fit the size of the business.<sup><a href=\"#src-s17\">17</a></sup> EU food law requires every business to know its suppliers and its business customers.<sup><a href=\"#src-s19\">19</a></sup> Traceability explains these rules.\n\nThe release decision follows the checks (our summary). A batch within its limits is released. A batch outside is held, and the cause is corrected.<sup><a href=\"#src-s1\">1</a>,<a href=\"#src-s17\">17</a></sup> For acrylamide, results that are not below the benchmark trigger a review of the mitigation measures without delay.<sup><a href=\"#src-s18\">18</a></sup>\n\n<!-- deep-dive -->\n\n## Reference values\n\n| Parameter | Value or rule | Source |\n|---|---|---|\n| Cupping roast, colour target (SCA) | Ground coffee L\\* 26–29, ≈ tile #65; batches outside are repeated | SCA 102-2024<sup><a href=\"#src-s1\">1</a></sup> |\n| Roast-meter conditions | Grind level and sample temperature controlled; much chaff gives a larger deviation | SCA 102-2024<sup><a href=\"#src-s1\">1</a></sup> |\n| Meter-specific cupping targets | Listed per meter brand (Table 4); scales arbitrary, brands do not fully correlate | SCA CVA document<sup><a href=\"#src-s2\">2</a></sup> |\n| Whole vs ground L\\*, arabica (D65) | 45.2 vs 49.8 (lightest); 40.0 vs 41.9; 37.1 vs 37.4 (darkest) | Bicho et al. 2012<sup><a href=\"#src-s4\">4</a></sup> |\n| Whole vs ground L\\*, robusta (D65) | 45.8 vs 52.5 (lightest); 40.5 vs 45.7; 38.2 vs 39.8 (darkest) | Bicho et al. 2012<sup><a href=\"#src-s4\">4</a></sup> |\n| Roast loss, typical | 15–22%; 14–23% | Münchow et al. 2020<sup><a href=\"#src-s3\">3</a></sup>, Bicho et al. 2012<sup><a href=\"#src-s4\">4</a></sup> |\n| Moisture of roasted coffee (basis not stated) | 1–2.5% | Münchow et al. 2020<sup><a href=\"#src-s3\">3</a></sup> |\n| Moisture, routine method | Loss in mass at 103 °C, roasted ground coffee; best on degassed coffee | ISO 11294<sup><a href=\"#src-s5\">5</a></sup> |\n| Water, maximum (national examples) | 5% by mass | Swiss beverage ordinance<sup><a href=\"#src-s7\">7</a></sup>; Austrian Food Codex<sup><a href=\"#src-s8\">8</a></sup> |\n| Charred beans, maximum (national example) | 1% by mass | Swiss beverage ordinance<sup><a href=\"#src-s7\">7</a></sup> |\n| Discoloured beans and foreign bodies (“Einwurf”), maximum (national example) | 2% by mass | Austrian Food Codex<sup><a href=\"#src-s8\">8</a></sup> |\n| Control limits, univariate chart | Usually mean ± 3 standard deviations (“3-sigma limits”) | NIST/SEMATECH<sup><a href=\"#src-s9\">9</a></sup> |\n| EU acrylamide benchmark, roast coffee | 400 µg/kg | Reg. (EU) 2017/2158<sup><a href=\"#src-s18\">18</a></sup> |\n| Acrylamide analysis, limit of quantification | ≤ 50 µg/kg (benchmark ≥ 125 µg/kg); limit of detection three tenths of it, ≤ 15 µg/kg (our calculation) | Reg. (EU) 2017/2158, Annex III<sup><a href=\"#src-s18\">18</a></sup> |\n| Acrylamide analysis, recovery | 75–110% | Reg. (EU) 2017/2158, Annex III<sup><a href=\"#src-s18\">18</a></sup> |\n| Acrylamide sampling frequency | At least annually for products with a known, well-controlled level; more often where the benchmark may be exceeded | Reg. (EU) 2017/2158, Annex III<sup><a href=\"#src-s18\">18</a></sup> |\n\n## Measuring roast degree in production\n\n**Colour meters.** SCA 102-2024 defines its cupping roast by the lightness of ground coffee, and in roast-meter terms as a reading of 63 on the Agtron “Gourmet” scale or its equivalent on other meters.<sup><a href=\"#src-s1\">1</a></sup> The CVA document lists those targets by meter brand and warns that the scales are arbitrary.<sup><a href=\"#src-s2\">2</a></sup> A reading therefore belongs to one instrument. A roastery that uses two meters needs its own comparison between them (our note). Grind, sample temperature and chaff all shift readings.<sup><a href=\"#src-s1\">1</a></sup>\n\n**Whole and ground.** In the 2012 study, arabica whole beans read L\\* 45.2 against 49.8 ground at the lightest of three roasts, and 37.1 against 37.4 at the darkest.<sup><a href=\"#src-s4\">4</a></sup> Robusta showed larger gaps at each roast.<sup><a href=\"#src-s4\">4</a></sup> The authors link the difference to the temperature gradient across the bean.<sup><a href=\"#src-s4\">4</a></sup> Roast degree and colour shows the full data.\n\n**Other indices.** SCA 102-2024 accepts dry-matter loss or volume increase as alternatives, once each is correlated with colour or spectrometry.<sup><a href=\"#src-s1\">1</a></sup> Roast loss is easy to log for every batch, but it also reflects the moisture of the green coffee (see Roast loss). Bulk density depends on roasting conditions as well as on the green coffee.<sup><a href=\"#src-s6\">6</a></sup>\n\n**Colour and flavour.** Two sensory studies that varied both colour and roast time found colour the stronger predictor of flavour.<sup><a href=\"#src-s3\">3</a></sup> It supports measuring colour on every batch (our note). Colour alone does not show how the roast reached its end point; the CVA document knows of no simple test for too-fast or too-slow roasts and relies on trained cuppers.<sup><a href=\"#src-s2\">2</a></sup>\n\n## Moisture of roasted coffee\n\nISO 11294 is the routine method for moisture in roasted ground coffee: the loss in mass at 103 °C.<sup><a href=\"#src-s5\">5</a></sup> Roasted coffee also loses carbon dioxide and other volatiles on heating. The standard is therefore most suited to degassed coffee.<sup><a href=\"#src-s5\">5</a></sup> A sample taken straight from the cooler includes gas loss in its reading (our reading of the scope).\n\nThe 1–2.5% reported for roasted beans is well below the 5% maximum in the Swiss and Austrian examples.<sup><a href=\"#src-s3\">3</a>,<a href=\"#src-s7\">7</a>,<a href=\"#src-s8\">8</a></sup> Water quenching can push coffee towards it; Cooling and quenching gives the data.\n\n## Statistical process control\n\nA univariate SPC chart plots one quality value per batch over time. It has a centre line at the mean of the in-control process.<sup><a href=\"#src-s9\">9</a></sup> Upper and lower control limits are set so that almost all points fall between them while the process stays in control.<sup><a href=\"#src-s9\">9</a></sup> The limits are usually three standard deviations from the mean.<sup><a href=\"#src-s9\">9</a></sup>\n\nColour, roast loss and moisture are single numbers per batch, so they suit such charts (our note). The limits come from the roastery's own data, not from a standard. No source consulted for this entry gives control limits for roast colour or roast loss.\n\n**Multivariate monitoring.** A 2018 study took the idea into the roast itself.<sup><a href=\"#src-s10\">10</a></sup> It recorded one near-infrared (NIR) spectrum per minute in a pilot drum roaster of about 250 g.<sup><a href=\"#src-s10\">10</a></sup> Fifteen batches from an experimental design built the model; five more carried deliberate faults.<sup><a href=\"#src-s10\">10</a></sup> The faults were a different blend, higher or lower power, and a lighter or heavier charge.<sup><a href=\"#src-s10\">10</a></sup>\n\nThe model used principal component analysis on a sliding window of 4 min with three components.<sup><a href=\"#src-s10\">10</a></sup> Two multivariate charts, Hotelling’s T² and the Q-residual chart (“squared predicted error”), flagged all five faults.<sup><a href=\"#src-s10\">10</a></sup> Timing differed. The wrong blend showed from the start. Low power appeared late, at the last two points of the T² chart, and the heavier charge crossed the T² limit only once.<sup><a href=\"#src-s10\">10</a></sup> The authors note that conventional control, by the senses of an expert, detects disturbances only at the end, if at all.<sup><a href=\"#src-s10\">10</a></sup>\n\n**In-line prediction.** A 2016 study from the same group placed an NIR probe in the roaster.<sup><a href=\"#src-s11\">11</a></sup> It predicted sucrose and colour during roasting of arabica and robusta from four origins, roasted slow and fast.<sup><a href=\"#src-s11\">11</a></sup> All calibrations reached a “range error ratio” (RER) above 10 and a coefficient of determination above 0.85.<sup><a href=\"#src-s11\">11</a></sup> Both studies are at pilot scale.\n\n## Uneven roasts and bean-to-bean variation\n\nA batch can match its target colour on average and still be uneven. One national text, the Austrian Food Codex, describes a correct roast as evenly brown and brown inside.<sup><a href=\"#src-s8\">8</a></sup> It notes that roasted coffee is usually sorted to remove discoloured beans and foreign bodies, and caps what remains at 2% by mass.<sup><a href=\"#src-s8\">8</a></sup> The Swiss ordinance caps charred beans at 1%.<sup><a href=\"#src-s7\">7</a></sup>\n\nAccording to the CVA document, cuppers can be trained to spot roasting problems from the beans’ structure, gradient of colours and flavour.<sup><a href=\"#src-s2\">2</a></sup> A ground-coffee reading averages over many beans. A 2022 imaging study states that coffee batches are not homogeneous.<sup><a href=\"#src-s12\">12</a></sup> It predicted aroma compounds in single roasted beans with hyperspectral imaging at 1000–2500 nm.<sup><a href=\"#src-s12\">12</a></sup> The authors see possible uses in standardising roast quality and removing defect beans; these were not tested in production.<sup><a href=\"#src-s12\">12</a></sup>\n\n## Sensory release checks\n\n**Which test.** ISO 4120 and ISO 10399 detect whether a perceptible difference exists, not how large it is or in which attribute.<sup><a href=\"#src-s14\">14</a>,<a href=\"#src-s15\">15</a></sup> The triangle test is statistically more efficient than the duo-trio test.<sup><a href=\"#src-s14\">14</a></sup> Its use is limited for products with strong carryover or lingering flavours.<sup><a href=\"#src-s14\">14</a></sup> The duo-trio test is easier for assessors. Its constant-reference form suits a familiar product, such as a sample from regular production.<sup><a href=\"#src-s15\">15</a></sup>\n\n**Homogeneous samples.** ISO 4120 applies only to homogeneous products.<sup><a href=\"#src-s14\">14</a></sup> Cups of coffee from one batch must therefore be brewed alike, and the batch sample must be mixed before it is split (our note). Sensory science methods gives the numbers of correct answers needed.\n\n**Cupping.** Cupping judges quality on a scale and depends on expert cuppers.<sup><a href=\"#src-s16\">16</a></sup> No study consulted for this entry measured how well production cupping detects batch deviations.\n\n## Food-safety monitoring\n\n**HACCP.** Regulation (EC) No 852/2004 requires procedures based on hazard analysis and critical control points (HACCP).<sup><a href=\"#src-s17\">17</a></sup> Operators identify hazards, set critical limits that separate acceptable from unacceptable, and set corrective actions.<sup><a href=\"#src-s17\">17</a></sup> Which steps of a roastery are critical control points is decided in each plant's hazard analysis. No source consulted for this entry assigns them for coffee roasting.\n\n**Foreign bodies.** The EPA describes destoners in roasting plants as air classifiers. They remove stones, metal fragments and other waste missed by the screening of green coffee.<sup><a href=\"#src-s20\">20</a></sup> The Austrian Codex includes foreign bodies in its 2% “Einwurf” limit.<sup><a href=\"#src-s8\">8</a></sup>\n\n**Acrylamide.** Regulation (EU) 2017/2158 asks coffee roasters to identify the critical roast conditions for minimal acrylamide within the target flavour profile.<sup><a href=\"#src-s18\">18</a></sup> Control of roast conditions belongs in a prerequisite programme under good manufacturing practice.<sup><a href=\"#src-s18\">18</a></sup> Blend recipes should take into account that robusta products tend to carry more acrylamide.<sup><a href=\"#src-s18\">18</a></sup>\n\nAnnex III sets the monitoring. Each product type needs a representative sample, and the sample must represent its batch.<sup><a href=\"#src-s18\">18</a></sup> Products with a known, well-controlled level are tested at least once a year; those that may exceed the benchmark more often.<sup><a href=\"#src-s18\">18</a></sup> For a benchmark of 125 µg/kg or more, the limit of quantification must be 50 µg/kg or lower.<sup><a href=\"#src-s18\">18</a></sup> The limit of detection is three tenths of it, so at most 15 µg/kg (our calculation).<sup><a href=\"#src-s18\">18</a></sup> Recovery must lie between 75% and 110%.<sup><a href=\"#src-s18\">18</a></sup>\n\n<figure>\n<div data-figure=\"roast-quality-control-acrylamide-analysis\"></div>\n<figcaption>What an acrylamide analysis of roast coffee must resolve under Regulation (EU) 2017/2158: limit of detection (three tenths of the limit of quantification; our calculation), limit of quantification and the benchmark level. Results not below the benchmark trigger a review of mitigation measures.<sup><a href=\"#src-s18\">18</a></sup></figcaption>\n</figure>\n\nRules as consulted on 16 November 2026. The legal text in force applies.\n\n## Risks and limits\n\n- **One instrument, one scale.** Colour readings from different meters are not interchangeable.<sup><a href=\"#src-s2\">2</a></sup> A replaced or recalibrated meter can shift every target.\n- **Averages hide spread.** Ground-coffee colour and batch roast loss are averages; scorched or underdeveloped beans can sit inside an on-target batch.<sup><a href=\"#src-s2\">2</a>,<a href=\"#src-s12\">12</a></sup>\n- **Green coffee drifts.** Roast loss reflects the moisture of the green coffee as well as the roast (see Roast loss). A change of green lot can move it with no change in roast degree.\n- **Fresh samples.** Moisture by loss in mass is best on degassed coffee.<sup><a href=\"#src-s5\">5</a></sup>\n- **Late detection.** Checks after cooling find faults only after the batch is finished.<sup><a href=\"#src-s10\">10</a></sup>\n\n## Open questions\n\n- **Control limits for roast colour.** No source consulted for this entry reports how much batch-to-batch colour variation production roasters achieve, or what tolerance is perceptible in the cup.\n- **Inline methods at scale.** The NIR studies used a pilot roaster of about 250 g.<sup><a href=\"#src-s10\">10</a>,<a href=\"#src-s11\">11</a></sup> No study consulted for this entry tested them on production roasters.\n- **Single-bean inspection.** Hyperspectral imaging of single beans is a laboratory method so far (2022).<sup><a href=\"#src-s12\">12</a></sup> Its use for sorting or release in roasteries is not established.\n- **Sampling plans.** No standard consulted for this entry sets how many samples per batch, or how often, a roastery should measure colour, roast loss or moisture.\n\n## Key studies\n\n<div class=\"study\">\n<p><strong>Catelani et al., 2018 – NIR and multivariate charts during the roast.</strong> Records near-infrared spectra each minute in a pilot drum roaster and builds a principal component model from fifteen designed batches. Five batches with a wrong blend, changed power or changed charge were all flagged by Hotelling’s T² and Q-residual charts, though some only late or weakly.</p>\n<p class=\"meta\">Talanta 179, 292–299 · <a href=\"https://doi.org/10.1016/j.talanta.2017.11.010\">doi:10.1016/j.talanta.2017.11.010</a></p>\n</div>\n\n<div class=\"study\">\n<p><strong>Santos et al., 2016 – In-line NIR for sucrose and colour.</strong> Places a near-infrared probe in the roaster and calibrates it, by partial least squares regression, against sucrose and colour. Arabica and robusta from four origins, roasted slow and fast, were predicted with a “range error ratio” above 10 for all parameters.</p>\n<p class=\"meta\">Food Chemistry 208, 103–110 · <a href=\"https://doi.org/10.1016/j.foodchem.2016.03.114\">doi:10.1016/j.foodchem.2016.03.114</a></p>\n</div>\n\n<div class=\"study\">\n<p><strong>Münchow et al., 2020 – Roast colour and time in the cup.</strong> Combines sensory studies that varied roast colour and roast timing. Colour predicted flavour better than timing; the authors note that specialty roasteries often leave colour out of QC.</p>\n<p class=\"meta\">Beverages 6(2), 29 · <a href=\"https://doi.org/10.3390/beverages6020029\">doi:10.3390/beverages6020029</a></p>\n</div>\n\n<div class=\"study\">\n<p><strong>Bicho et al., 2012 – Colour of whole and ground beans.</strong> Roasts one arabica and one robusta at three intensities and measures colour of whole and ground beans. Whole beans read darker, and the gap narrowed as the roast got darker.</p>\n<p class=\"meta\">Ciência e Tecnologia de Alimentos 32(3), 436–442 · <a href=\"https://doi.org/10.1590/S0101-20612012005000068\">doi:10.1590/S0101-20612012005000068</a></p>\n</div>\n\n<div class=\"study\">\n<p><strong>Caporaso et al., 2022 – Aroma of single beans by imaging.</strong> Uses hyperspectral imaging at 1000–2500 nm to predict volatile compounds in single roasted beans. The authors stress that batches are not homogeneous and propose the method for standardising roast quality and removing defect beans.</p>\n<p class=\"meta\">Food Chemistry 371, 131159 · <a href=\"https://doi.org/10.1016/j.foodchem.2021.131159\">doi:10.1016/j.foodchem.2021.131159</a></p>\n</div>"
}