Short answer
A data sheet quotes two figures, repeatability and inter-instrument agreement, and both are best cases: repeat readings of a white tile, and the average difference between instruments of one model over a set of ceramic tiles. Neither covers agreement between different models, which is far worse, or behaviour on real samples. A fleet is matched by standardising the model and settings, verifying each instrument against shared physical standards, and applying profiling corrections, and the tolerance is then set above whatever disagreement remains.
Repeatability is how closely one instrument agrees with itself on one sample over a short time. Inter-instrument agreement, as makers use the term, is how closely two instruments of the same make and model agree. Inter-model agreement, between different designs, is the third figure, and it does not appear on data sheets. ASTM's practice for specifying instrument performance exists to give these terms fixed meanings and test procedures; it supplies the format for a specification and deliberately sets no values. The distinction matters commercially. A supplier and customer who each own an excellent instrument of different models have bought two good repeatabilities and no stated agreement with each other at all.
The table sets out three published specifications. Repeatability is quoted on the white calibration tile: a maximum of 0.01 over twenty readings for one benchtop instrument, a standard deviation within 0.02 over thirty readings for a portable one. Agreement is quoted as an average: 0.08 for the benchtop, with a maximum of 0.15, and 0.12 or 0.2 for two portables, taken over twelve ceramic tiles against a master instrument with the specular component included. The conditions are part of the figure: one sheet ties its numbers to 23 °C and 44 per cent humidity and to annual calibration. Each choice flatters the result legitimately. Tiles are flat, glazed, opaque and stable; an average conceals the worst tile; and a white tile is the easiest thing an instrument ever measures.
Set those figures beside what happens between designs. The contract-proofing standard in print treats ΔE00 limits below 2.5 as impractical because instruments of different models agree so poorly, and recommends halving its tolerances when a single instrument makes both measurements. That floor is thirteen to thirty-one times the same-model averages in the table. The comparison is rough, since the units and materials differ, but the order of magnitude is the lesson: different optics, apertures, lamps and ultraviolet content make different systematic errors on the same sample. A study of ten graphic-arts instruments found the gap between models narrowed sharply once all measured to the same defined ultraviolet condition, which shows how much of it comes from undefined settings.
Fleet matching has three layers. First, remove the differences that are choices: one model where possible, and identical geometry, specular mode, aperture, illuminant and observer everywhere. Second, verify each instrument regularly against stable physical standards. The usual set is twelve glazed ceramic tiles, three greys for linearity, seven colours for spectral response and two colour-difference pairs, each with calibrated values. Third, profile: software compares each instrument's readings of such a set with reference values and applies a correction, and flags an instrument that has drifted far enough to need service. One maker claims an average improvement of 0.2 ΔE* for one family of hand-held instruments. Profiling corrects systematic differences on tiles. It cannot make two geometries equivalent.
The figure that matters for a contract is none of the above; it is how far the two parties' instruments disagree on the actual product. That has to be measured. Circulate a set of stable tiles and a set of real samples spanning the colour range, including the most saturated, the darkest and the most textured. Have each site measure them by the written procedure, several times with repositioning. The spread between sites on the tiles shows instrument disagreement; the extra spread on real samples shows what presentation and conditioning add. A tolerance has to leave room for both, and when a tighter limit is truly needed the remedy is a single referee instrument, not a smaller number.
| Instrument | Type | Repeatability, as stated | Inter-instrument agreement, as stated | ISO 12647-7 inter-model floor ÷ stated average |
|---|---|---|---|---|
| Datacolor Spectro 1000 series | Benchtop d/8° sphere | 0.01 maximum (CIELAB) — 20 readings of the white tile, double flash | 0.08 average, 0.15 maximum (CIELAB) — Reflectance, at 23 °C ± 1 °C and 44 % ± 1 % relative humidity; annual calibration required | about 31× |
| Konica Minolta CM-26dG and CM-26d | Portable d/8° sphere | Standard deviation within ΔE*ab 0.02 — White calibration plate, 30 readings at 5-second intervals | Within ΔE*ab 0.12 (average) — Average over 12 BCRA Series II tiles, medium aperture, specular included, against a master body | about 21× |
| Konica Minolta CM-25d | Portable d/8° sphere | Standard deviation within ΔE*ab 0.04 — White calibration plate, 30 readings at 5-second intervals | Within ΔE*ab 0.2 (average) — Average over 12 BCRA Series II tiles, medium aperture, specular included, against a master body | about 13× |
Why: Different aperture, specular mode or illuminant and observer settings, or one instrument overdue for service.
Fix: Compare the settings line by line, then measure a shared tile set on both.
Why: The product is textured, translucent or temperature-sensitive, and the sites present it differently.
Fix: Write and audit the sample-presentation procedure; the instruments are not the problem.
Why: A different design makes different systematic errors, particularly on dark, saturated or brightened samples.
Fix: Run old and new in parallel on real samples, and re-derive or re-reference tolerances before switching.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
StandardStrong evidence
ASTM E2214-23 provides standard terminology and procedures for characterising the performance of colour-measuring instruments, giving the format by which specifications are determined, communicated and verified without setting the specifications themselves.
FactStrong evidence
One benchtop sphere spectrophotometer is specified at a repeatability of 0.01 maximum over 20 readings of a white tile and an inter-instrument agreement of 0.08 average and 0.15 maximum in CIELAB units, at 23 °C ± 1 °C and 44 % ± 1 % relative humidity, with annual calibration required.
Caveat: A manufacturer's specification for one product line; not a measured result on production samples.
FactStrong evidence
One maker specifies its portable sphere spectrophotometers at an inter-instrument agreement within ΔE*ab 0.12 or 0.2 depending on model, as an average over 12 BCRA Series II tiles with specular included against a master body, and a repeatability standard deviation within 0.02 or 0.04 over 30 readings of the white calibration plate.
Caveat: A manufacturer's specification under its own standard conditions.
Source: Konica Minolta to Release New Portable Spectrophotometers (CM-26dG, CM-26d, CM-25d)
Colourwise analysisLimited evidence
The 2.5 ΔE00 that ISO 12647-7:2016 treats as the practical floor between instruments of different models is roughly 13 to 31 times the average same-model agreement stated in three published instrument specifications.
Based on: Colourwise divided the ISO figure by each maker's stated average inter-instrument agreement; see the table on this page.
Caveat: An order-of-magnitude comparison only: the ISO figure is in ΔE00 on printed proofs with 45°/0° instruments, the specifications are in ΔE*ab on ceramic tiles with sphere instruments.
Source: ISO 12647-7:2016 … — Part 7: Proofing processes working directly from digital data; Datacolor Spectro 1000 series specification sheet; Konica Minolta to Release New Portable Spectrophotometers (CM-26dG, CM-26d, CM-25d)
FactStrong evidence
The Ceramic Colour Standards Series II set used to verify colour instruments consists of twelve glazed ceramic tiles: three neutral greys, seven chromatic standards and two colour-difference standards.
FactLimited evidence
One maker's profiling service validates instruments against traceable physical colour standards and claims an average improvement of 0.2 ΔE* for one family of hand-held instruments.
Caveat: A vendor's claim on a product page, with no method or sample stated.
ConventionModerate evidence
Instrument makers define inter-instrument agreement as agreement between spectrophotometers of the same make and model, and warn against mixing instrument geometries or models when measurements are to be compared.
Source: Why Benchtop Spectrophotometer Inter-Instrument Agreement & Repeatability Matter for Brand Suppliers
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.