CRI and Colour Rendering: What Ra and R9 Really Tell You

Two light sources can share the same colour temperature and yet make the same room, the same painting or the same face look entirely different. Colour rendering — the ability of a source to reveal surface colours faithfully — is where that difference lives. This guide explains what the CRI number on a datasheet actually measures, where it misleads, and how to specify colour quality in a way that survives contact with a real project.

What CRI actually measures

The Colour Rendering Index compares how a test source renders a set of standard colour samples against how a reference source of the same colour temperature renders them — an incandescent-style reference at warm colour temperatures, a daylight model at cooler ones. The headline figure, properly written Ra, is the average score across the first eight samples, R1 to R8, which are all relatively muted, pastel tones. A source that shifts those samples very little scores close to 100; larger shifts pull the average down. An incandescent lamp scores essentially 100 by definition, ordinary commodity LED sits around 80, and quality architectural LED reaches the mid 90s.

The key word is average. Ra says nothing about which colours a source renders poorly, only how it performs across eight gentle tones taken together. Two sources with identical Ra can fail in different regions of the colour space — and the region where LED most often falls short is exactly the one the standard eight samples avoid: saturated red.

Why R9 matters

R9 is the score for a deep, saturated red sample that sits outside the Ra average. It matters because saturated red is structural to how interiors read. Skin tones depend on red rendering — blood beneath skin is what gives a complexion warmth and life, and a low-R9 source leaves faces grey and drawn. Food, timber, leather, terracotta, and a very large proportion of pigments used in paintings and textiles all lean on red content. An LED can score Ra 90 while managing an R9 in the teens, and the difference is visible to a lay observer even if they cannot name it: the room simply feels less alive. This is why serious specifications always state R9 alongside Ra, and why galleries, restaurants and prime residential projects treat it as non-negotiable.

CRI 80 versus CRI 90 in practice

CRI 80 is the general commercial baseline — acceptable for back-of-house, utility and many workplace applications where colour fidelity is not the point of the space. The step to CRI 90 and above is clearly perceptible in any interior where materials, finishes, art or people are meant to be seen well: joinery gains depth, stone veining reads correctly, and skin tones warm up. The efficacy penalty for higher CRI — a few per cent more power for the same light — was a genuine trade-off a decade ago and is now marginal. In residential work there is rarely a good reason to accept CRI 80 anywhere a person will linger.

The limits of CRI — and TM-30

CRI has known weaknesses: eight desaturated samples are a thin basis for judging a modern LED spectrum, an average conceals individual failures, and a source can be tuned to score well on the test samples without rendering real materials faithfully. The IES TM-30 method was developed to address this. It evaluates 99 colour samples drawn from real objects and reports two headline values: Rf, a fidelity index on a similar 0–100 basis, and Rg, a gamut index describing whether the source tends to mute colours (Rg below 100) or saturate them (Rg above 100). TM-30 is increasingly quoted by better manufacturers and is worth requesting for colour-critical spaces, but CRI plus R9 remains the common contractual language of specification, so the practical approach is to specify Ra and R9 firmly and use TM-30 data as supporting evidence where it is available.

Consistency between fittings: SDCM and MacAdam ellipses

Colour rendering describes how a source treats objects; colour consistency describes whether the sources themselves match one another. LED chips vary slightly in manufacture, so they are sorted into bins, and the tightness of a bin is expressed in SDCM — standard deviation of colour matching — based on MacAdam ellipses, which map the smallest colour differences the eye can detect. Within 1 SDCM, no difference is visible; at 2, a trained eye may notice; at 3, differences are on the threshold of visibility for most people; by 5 or more, adjacent fittings on the same white ceiling can look plainly mismatched. Architectural-grade product is typically binned at 3 SDCM or better, and premium ranges at 2. The failure mode to avoid is mixing batches or ranges with loose binning across one visual field — a row of wall-washers is unforgiving of even small colour differences.

Practical specification advice

For prime residential and hospitality interiors, the working baseline is CRI (Ra) of 90 or above with R9 of at least 50, and 3 SDCM or tighter colour consistency across all fittings that share a sightline. For galleries, dressing areas and anywhere colour judgement matters, push R9 higher — many quality ranges now achieve R9 above 90. Writing these values into the schedule, and checking submittals against them, is a routine part of proper product specification; the table below summarises sensible minimums by application.