Color Reading and Measurement

Color measurement is a useful roasting metric when it is treated as a controlled measurement system, not as a universal roast-quality score. This page explains how Roest users use color meters for QC, profile comparison, whole-bean versus ground-color analysis, and troubleshooting measurement error. It also covers why readings often fail to transfer between grinders, devices, roast ages, and roasters.

What Color Measurement Is Good For

Color is most useful as a repeatability and benchmarking tool. It can confirm whether a known profile is landing near a previous successful roast, help compare whole-bean and ground development, and provide another data point alongside weight loss, drop temperature, development time, and cupping.

It is not a standalone definition of development. The same color can taste different when reached through different roast times, inlet shapes, pressure environments, or machines. One comparison found that matching color, grind size, brew time, and extraction yield between Roest and Kaffelogic still produced very different taste source. Another warning was that roasting a bean to the same color with first crack at 4, 6, or 10 minutes would not make the roasts taste the same source.

Color is therefore best used with sensory evaluation. If the graph and color are both the same, a roast is more likely to taste similar; graph alone is not enough source. For how to validate roast changes in the cup, see Cupping and Sensory Evaluation.

Standard Measurement Protocol

The most important rule is consistency. A color number is only comparable when the device, calibration, sample preparation, grind, roast age, and reading method are controlled.

1. Record the measurement context. Log the device, scale, grinder, burrs, grind setting, sample type, sample amount or filled volume, tamping/leveling method, roast age, and whether the reading is whole bean or ground. Comparisons without calibration, device, and sampling protocol are unreliable source.

2. Warm and clean the meter. One repeatable DIY-meter workflow keeps the meter on for 10 minutes before reading and cleans with alcohol and Q-tips between readings source. Glass cleanliness matters: readings can change after 5-10 samples if the glass is not cleaned source.

3. Check calibration before relying on results. For the community DIY meter, baking soda/sodium bicarbonate has been used as a calibration check around 230, and brown unbleached Hario/V60 paper stacks have been used around the mid-180s 2 sources. DIY meter builds are sensitive to glass thickness and sensor distance, so a mechanically different build can read differently even if the software is correct 2 sources.

4. Prepare ground samples the same way every time. For the DIY meter style described in the community, the grind is typically espresso-fine or finer, with enough coffee to fill the chamber and a light one-finger tamp; inconsistent tamping or escaped light makes readings inconsistent source. If the coffee clumps, adjust the grind or preparation rather than treating the number as reliable source.

5. Take multiple readings when the device and sample allow it. A common workflow is at least three readings, discarding obvious high or low outliers and averaging the remainder source. A one-point bean-color difference can fall within sample variance, so small differences should not drive profile decisions without sensory confirmation source.

6. For whole-bean readings, redistribute the sample. Repeated readings without stirring may not add information; in one workflow, multiple readings only helped after stirring the beans, and even then showed about ±5 variation source. Devices with vibrating platforms or distribution routines still need consistent fill and leveling source.

7. Measure at a consistent roast age. Color and grind behavior can shift after roasting, so QC comparisons should be taken at a consistent interval. If a production profile is judged at 24 hours, later batches should be measured at the same interval unless the purpose is specifically to study aging.

Quick Troubleshooting Table

Grind Size Is the Largest Hidden Variable

Grind size can move color readings enough to make two identical roasts appear different. In one sieved test, the same coffee read 114 Agtron under 250 µm, 101 Agtron at 250-500 µm, and 84 Agtron above 500 µm; the unsieved sample read 101 Agtron source. Another estimate was that grinding to an SCA-style range of 600-850 µm would read about 20 Agtron lower than a finer-than-espresso grind on the same calibrated meter source.

The community rule of thumb is simple: finer grind reads higher/lighter, coarser grind reads lower/darker source. This has been observed in specific grinder comparisons: Kinu settings produced roughly 117 versus 121/122 on the same coffee, and an EK43 SSP HU comparison gave 104 at dial 4 versus 112 at dial 1 2 sources.

Because burrs and grind distribution matter, changing a burrset can make previous reference points unusable source. A color protocol should therefore specify not only “espresso” or “filter,” but the grinder, burrs, setting, and, if available, a particle-size target.

Whole Bean, Ground Color, and Color Delta

Whole-bean color and ground color answer different questions. Whole-bean readings reflect surface color. Ground readings better represent interior development, though they are more sensitive to grind preparation. Comparing the two can reveal development gradients that are invisible by eye.

One Lighttells 2 reference for a well-developed light roast was a ground/whole-bean difference under 40 points with whole-bean color above 70 source. Another practical QC example used 70/110 as a successful profile reference; production batches with the same profile were expected around that same value within -2/+1 source.

Development time often changes the inside more than the outside. In one set of tests, reducing development to 30 seconds did not change outside end color, but made the inside lighter source. The same contributor summarized development time as the phase that closes the gap between bean surface and bean core in color measurement source.

A large delta is not automatically a defect. Attempts to force a very close whole-bean-to-ground spread can be associated with persistent dryness, especially when achieved through very high airflow or long roast duration source. Whole-bean color is also especially prone to measurement issues, and attempts to manipulate color delta can behave counterintuitively source.

How to Use Color in QC and Profiling

For production or repeated batches, color is strongest when tied to a known good cup. A profile that cups well at a given whole/ground color can use that number as a QC target for future runs. In one example, a successful 70/110 profile was expected to reproduce near 70/110, within -2/+1, on subsequent production batches using the same profile source.

For profiling, color helps identify the direction of adjustment rather than giving the whole answer. One example interpretation was that a 65/120 test roast suggested reducing late energy and extending Maillard and development time source. Another common pattern is that an outside-dark/inside-light result points toward excessive surface energy relative to core development; that diagnostic overlaps with Heat Transfer Fundamentals, Airflow and Fan Settings, and Development Time and Drop Decisions.

Color should not replace cupping. It can flag likely underdevelopment, roastiness, or inconsistency, but it cannot tell whether a roast is baked, whether it will age well, or whether a given profile preserved the desired aromatics. One contributor summarized the limitation clearly: color can be another indicator of similarity when the curve and drop are also aligned, but it cannot tell baked versus not baked source.

Comparing Numbers Across Devices and Scales

Color numbers are not portable unless the scale and method are identical. The same nominal value can mean different things on a DIY meter, Lighttells, Roast Vision, Tonino, Roastrite, DiFluid, Omix, KN-201, or Colortrack. Even when two users both report “Agtron,” device calibration and sample prep can produce different values.

Some scales also run in different directions. On the DIY/Lighttells/Agtron-style measurements commonly discussed in the Roest community, higher values generally mean lighter color. By contrast, Colortrack was explicitly described with higher numbers meaning darker roast source.

The following values are examples, not universal targets:

Because references are device-specific, shared targets should include the device and sample protocol. A 70/110 Lighttells reference does not mean another roaster should expect 70/110 on a different meter, grinder, roast age, or coffee.

DIY Meter Notes

The community DIY roast meter is based on the SparkFun MAX30101 sensor, a sensor family also mentioned in relation to several color meters including Roast Vision source. The build has been described as using SparkFun parts, a 3D-printed case, microscope glass, and no soldering, with parts plugged together using pre-made cables source.

Reported cost varies. Some described the parts as under or around $100, while another report put electrical parts above $120 2 sources. CONFLICT (Unresolved): the cost disagreement likely depends on which parts, region, shipping, case, and glass are included, but the claims do not provide enough detail to normalize the estimate.

For DIY builds, mechanical consistency matters. Standard microscope glass around 1.2 mm was specified in one build note, and the distance from the glass to the sensor was said to influence readings source. If a DIY meter reads a known calibration material correctly but coffee readings appear implausible, glass thickness, sensor distance, firmware version, and tamping/leveling should be checked before changing roast profiles.

Color and Roast Age

Color readings and apparent roast color can change after roasting, but the mechanism is disputed.

The practical conclusion is the same either way: measure at a consistent roast age, especially for QC. If comparing fresh, 24-hour, 7-day, and 30-day readings, label them separately and do not mix them as if they were the same measurement condition. For sensory timing, see Resting and Degassing.

Visual Color Is Not a Substitute for Measurement

Small numerical differences are not reliably visible. A 72 versus 75 or 69 versus 72 difference was called impossible to distinguish by eye source. Photos are even less reliable because tray color, background, exposure, and white balance can alter perceived roast color 2 sources.

Visual inspection is still useful for defects, bean expansion, tipping, chaff, and inside/outside contrast, especially when beans are cut open. For defect interpretation, see Roast Defects Troubleshooting. For repeatable numeric tracking, use a controlled meter protocol rather than photos or naked-eye comparisons.

Practical Rules

Use color measurement as a controlled instrument, not as a universal truth. The most robust workflow is to establish a house reference coffee, measure it with the exact grinder and meter protocol, cup it, then use that reference for relative comparisons.

Good color practice follows five rules:

1. Keep device, calibration, grind, sample prep, and roast age fixed.
2. Record whole and ground readings separately.
3. Use color with weight loss, drop temperature, roast curve, and cupping.
4. Re-baseline after changing grinder, burrs, meter, glass, firmware, or sample protocol.
5. Do not chase another roaster’s number unless their device, grind, sample preparation, and coffee context are known.

Color is most powerful when it answers a narrow question: “Did this roast land like the last version I liked?” It is much weaker when used to answer “Is this coffee good?” without taste, context, and process information.