CRI is a score from 0 to 100 that compares a lamp against an ideal source of the same colour temperature, using eight muted colour samples. In practice, asking what is CRI in lighting usually means asking whether a fixture is any good, which is a much larger question than the metric answers. What the score does tell you is how faithfully colours are rendered on average. It does not tell you whether reds look right, and on hospitality projects that is usually the only part of the question that matters. Short answer: specify Ra ≥ 90 and R9 ≥ 50 for guest areas, and stop treating a single CRI number as proof of quality.
1. What CRI actually measures
The colour rendering index was published by the CIE in 1965 and revised in 1995 as CIE 13.3. The method is a comparison, not an absolute measurement. A test source is compared against a reference source of the same correlated colour temperature — a blackbody radiator below 5000 K, a daylight model above it — and the visible differences between them are scored.
Two things follow from that definition, and both are routinely ignored. First, CRI measures fidelity, not appeal. A source that scores 100 matches the reference exactly, which for warm interiors means matching a glowing filament. That is not the same as making food look appetising or skin look healthy. Second, because the comparison is always at matched colour temperature, a CRI of 90 at 2700 K and a CRI of 90 at 4000 K do not describe the same spectrum, and the two numbers are not directly interchangeable.
What CRI does not measure: saturation, individual hue families, flicker, glare, uniformity, or how the light behaves when dimmed. A fixture can score Ra 95 and still strobe at 20% output. The metric is narrow, and it is only useful when you know how narrow.
2. How Ra is calculated from eight color samples
The calculation uses eight test colour samples, TCS01 to TCS08, chosen in the 1960s from the Munsell atlas. They are deliberately low in saturation: a dusty pink, a mustard, an olive green, a light blue, plus skin-adjacent and foliage-adjacent tones. Each sample is rendered under the test source and under the reference, and the colour difference ΔE in CIE 1964 U*V*W* is converted into an individual score R1 to R8.

| Sample | Approximate colour | What it proxies | Typical R value at Ra 80 |
|---|---|---|---|
| R1 | Light greyish red | Plaster, warm neutral paint | 78–82 |
| R2 | Dark greyish yellow | Oak, cane, linen | 80–85 |
| R3 | Strong yellow green | Foliage, moss | 74–80 |
| R4 | Moderate yellowish green | Olive fabric, sage paint | 76–82 |
| R5 | Light bluish green | Pool tile, teal upholstery | 73–79 |
| R6 | Light blue | Denim, grey-blue stone | 74–80 |
| R7 | Light violet | Cool grey, lavender | 76–82 |
| R8 | Light reddish purple | Mauve, rose marble | 72–78 |
Those eight scores are averaged and scaled. The 1995 version of the formula uses Ra = 100 − 4.6 × ΔE averaged across the samples, which is why the scale compresses at the top: a source with visible deficiencies still lands in the high 80s. There is no R9 in the average, and no sample darker than a mid-tone.
The bias is systematic. Every one of the eight samples is a light, desaturated object. Nothing in the set is a saturated red, nothing is a deep blue, nothing is a dark brown. So the metric is well suited to checking whether a white wall looks clean, and poorly suited to checking whether a walnut headboard or a plate of salmon looks right.
3. Why R9 is the number designers check
R9 is the ninth test sample: saturated red, drawn from the same atlas but excluded from the Ra average because a large group of 1960s lamps could not render it at all. It is reported separately, on the same 0 to 100 scale, and it is the single most useful number on a photometric report for interior work.
The reason is physical. Wood, leather, terracotta, tomato, beef, dark skin and most warm textiles depend on red reflectance above 600 nm. A spectrum with a red deficit renders all of them as grey-brown. Phosphor-converted white LEDs commonly show exactly that deficit, because the red phosphor is the expensive part of the mix.
| Reported CRI | R9 you will often see | How it reads in a room |
|---|---|---|
| Ra 80 | 0–15 | Timber goes flat, food looks tired, lips look grey under close light |
| Ra 90 | 25–50 | Acceptable on light finishes, weak on walnut, oxblood and rare meat |
| Ra 92 | 50–65 | Good general hospitality performance; most guests read it as natural |
| Ra 95 | 70–90 | Retail and restaurant standard; colour critical work is viable |
A steakhouse and a hotel bathroom prove the point faster than any table. A steakhouse with Ra 90 / R9 12 will serve beef that looks brown under the table light, and no amount of plating fixes it. A hotel bathroom with Ra 90 / R9 15 will make the mirror unflattering in a way guests notice without being able to name — the standard complaint is that the light “makes me look ill”.
For hospitality the working minimum is Ra ≥ 90 with R9 ≥ 50. For retail display and food service, R9 ≥ 70. Where the finish palette is dark timber or deep red upholstery, ask for an R9 figure against a specific sample rather than a generic claim, because the same fixture can score R9 45 on a lab sample and read worse on a lacquered surface.
R9 is not the whole colour story either. It is one saturated red. The saturations that matter for a scheme — the exact red of a velvet banquette, the green of a marble — are better covered by a full spectral report. That gap is what TM-30 addresses.
4. CRI 80 vs 90 vs 95 in real rooms
The step from 80 to 90 is obvious to anyone. The step from 90 to 95 is obvious to a designer and mostly invisible to a guest, unless the room is dark, the finishes are saturated, or the light is close to the surface.
| Kleurweergave-index (CRI) | Visible effect at 300 lx | Cost and availability | Use it for |
|---|---|---|---|
| 80 | Colours close but muddy; reds and greens separate poorly | Cheapest, widest choice, shortest lead time | Corridors, plant, car park, staff areas |
| 90 | Natural on most finishes; skin looks acceptable | Standard for hospitality, no real constraint | Guestrooms, bathrooms, public areas |
| 95 | Finishes read as themselves; deep reds and blues hold | Fewer options, longer lead time, higher BOM cost | Restaurants, bars, retail, art, vanity |
Where the budget is tight, spend it unevenly rather than everywhere. A guestroom at Ra 90 with a Ra 95 bedside reading light costs little more than an all-90 room and performs better where the guest is closest to the light. The reading position is 300 lx at 0.4 m from the page, and at that distance the spectrum is the only thing the eye has to judge.
The reverse also holds. Putting Ra 95 in a car park wastes money, because nobody evaluates colour at 100 lx from 3 m away. Match the metric to the viewing distance and the task.
5. CRI is not color temperature
These two get conflated more than any other pair in lighting. CRI describes how faithfully colours are rendered. CCT describes where the white itself sits, warm or cool. They are independent axes, and a fixture can be any combination of the two: warm and accurate, warm and poor, cool and accurate, cool and poor.

| Kleurtemperatuur | Character | Typical Ra requirement | Independent of CRI? |
|---|---|---|---|
| 2700 K | Warm, incandescent-like | Ra ≥ 90, R9 ≥ 50 | Ja |
| 3000 K | Warm white, halogen-like | Ra ≥ 90, R9 ≥ 50 | Ja |
| 3500 K | Neutral warm | Ra ≥ 85 | Ja |
| 4000 K | Neutral | Ra ≥ 80 (≥ 90 for inspection) | Ja |
Practical consequence: a warm room with poor CRI looks yellow and dirty, not cosy. A cool room with good CRI looks clinical but accurate. Neither problem is solved by moving the CCT. If the complaint is “everything looks grey”, check R9 first; if the complaint is “it feels cold”, check CCT and illuminance.
There is one interaction worth knowing. Because CRI is scored against a reference at the same CCT, low-CCT sources are compared against a very red reference, and high-CCT sources against a bluer one. That makes Ra slightly easier to achieve at 2700 K than at 4000 K for the same phosphor package, which is one reason premium 4000 K product is harder to source than premium 2700 K.
6. What to write in a specification
Write CRI as two numbers with a test method, not as an adjective. “High CRI” is not a specification, and neither is “CRI > 90” without an R9 value and a report reference.
| Area | Clause | Evidence required |
|---|---|---|
| Guestroom, bathroom | Ra ≥ 90, R9 ≥ 50, CCT 2700 K or 3000 K ± 150 K | LM-79 report per SKU, sample approval |
| Restaurant, bar | Ra ≥ 92, R9 ≥ 60, CCT 2700 K | LM-79 plus TM-30 Rf ≥ 90 |
| Retail display | Ra ≥ 95, R9 ≥ 70, CCT 3000 K | LM-79 plus TM-30 with vector graphic |
| Back of house | Ra ≥ 80, R9 not specified | LM-79 report |
Then add the three clauses that stop the number being gamed. Reports must be issued by an ISO/IEC 17025 accredited laboratory. They must be less than 24 months old. They must describe the complete luminaire — LED board, optic, diffuser, driver — not a bare board measured on a bench.
Finally, sample what arrives. One fixture in twenty, measured in the same conditions as the report. If the measured R9 is below the specified value, the batch is non-compliant regardless of what the Ra says, and the remedy is replacement rather than a credit note.
The fuller version of this framework, including TM-30 and SDCM, is in the guide to light quality metrics. Photometric files and test reports for a live project are available through the testing and documentation service.
Frequently asked questions
What is a good CRI for hotel lighting?
Ra 90 with R9 at 50 or above for guestrooms and public areas. Restaurants and bars should go to Ra 92 and R9 60, because food and dark timber are judged at close range. Back-of-house areas are fine at Ra 80, where nobody evaluates colour.
Is CRI 80 bad?
Not bad, just limited. At Ra 80 most finishes read as themselves at a glance, but saturated reds collapse and skin looks dull. It is the right choice for corridors, plant rooms and car parks, and the wrong choice for anywhere a guest looks closely at a surface.
Can CRI be higher than 100?
No. The scale is anchored so that a source identical to the reference scores 100. Values above 100 appear in some older datasets where measurement tolerances pushed the result, but they are not meaningful. Anything reported above 100 should be treated as 100 with measurement noise.
Does higher CRI always mean better light?
No. CRI measures fidelity against a reference, not comfort, glare, flicker or uniformity. A Ra 95 downlight with a 60° beam and no baffle can be worse in a guestroom than a Ra 90 fixture with proper shielding. Judge the whole fixture, not the score.
Next step
If the specification is already written, send it and the team will confirm which SKUs meet it and where an R9 value needs to be tested rather than assumed. For CRI 90+ downlights, the downlight range lists the tested values per family.






