What Does CRI 98 Mean on an LED Light, and Where Is R9?

CRI 98 is the figure given for this light, but even a true 98 tells you very little about deep red rendering. Here is what the number does and does not mean. The headline CRI figure is an average of eight muted test swatches, R1 to R8, and the saturated red sample buyers actually care about is R9, which sits outside that average altogether.

So a source can score 98 and still leave a terracotta bowl looking brown. This fitting's color rendering is recorded as CRI 98, with no R9 value or test report behind it.

Where does the CRI 98 figure on this page come from?

It's rated CRI 98. The same table gives 7W to 20W, three colour temperatures at 3000K, 4000K and 6000K, AC85V-260V, and a 24 degree beam angle also carried from source. No R9 value appears on it, and no fidelity and gamut pair.

That gap matters because reported colour numbers do drift from measured ones. The US Department of Energy's CALiPER programme ran a batch of OLED luminaires through outside laboratory work and found reported CCT and Ra within 4 per cent of test values in almost every case, with one exception: the maker of the unit logged as CALiPER 16-18 "claimed 90 CRI (Ra) while the test data show 79 CRI (Ra)". Eleven points, on a branded product with a spec sheet.

A CRI figure with no R9 beside it and no report behind it is a sales claim, not a measurement. None of that makes 98 false. It makes it unchecked, which is a different thing and the only honest label for it. Our note on whether an LED flickers at the lowest brightness setting hits the same wall: some qualities of a light source are never published, so you end up testing the one in your hand.

What does CRI 98 on an LED light mean, and what does it leave out?

It means a source matched a reference across eight pastel swatches, averaged into one score, and that is the whole of it. The Illuminating Engineering Society defines the index as "a measure of the degree of color shift that objects undergo when illuminated by the light source, as compared with the color of those same objects when illuminated by a reference source of comparable color temperature". SLS Lighting gives the arithmetic: the CIE "established a scale of 8 CIE Standard Colour Samples for the CRI method (R1-R8)", and "Ra is the average of the 8 Ri values and is stated as CRI with a maximum value of 100".

Seven further samples exist and none of them count. Extended CRI covers "the remaining 7 colours (9 to 15) including the saturated colours and skin tones", while the score on the box stays built from R1 to R8. Samples carry Munsell notation, where the final figure is chroma — how saturated the swatch is.

The eight samples that build a 98 run from chroma 4 to chroma 8. TCS09, strong red, is chroma 13, nearly twice the saturation of the most colourful swatch in the average. The hardest colour for an LED spectrum to render is the one the headline score never looks at.

Why do reds still look muddy under a light rated CRI 98?

Because the sample that scores saturated red is not part of the score. R9 "represents how accurately a light source will reproduce strong red colours", and SLS Lighting adds that it is "of special interest because skin or flesh tones are represented by this colour". It sits outside Ra, so a spectrum with a hole where red belongs can average well across the pastels and still flatten a terracotta bowl.

The shortfall can run past zero. The special indices were formulated so a value could come out negative, and strong red is where that happens. CALiPER's OLED report shows one: a luminaire measuring Ra 78 whose "R9 value is below zero", with a gamut illustration that "suggests that it would desaturate reds" and red-hue-bin fidelity nine points below the other panels tested. Ra 78 sounds respectable. The light drains red.

Regulators worked this out years ago. California's Title 24 sets a floor of R9 50 for homes, written because Ra does not reach that far. So if roast meat reads grey and a rust-coloured pot reads brown under a fitting claiming 98, your eyes and the number are not arguing — the number was never scoring that colour, which is the same collision our piece on whether the rust brown disc looks orange or red in person runs into.

From the shop

Adjustable Ceiling Spotlight, Surface Mounted LED Downlight, CRI 98

It's rated CRI 98 with a 24 degree beam angle; no R9 figure is given. Surface mounted, 7W to 20W, in 3000K, 4000K or 6000K on AC85V-260V.

View the adjustable ceiling spotlight — 6 fixture colours & current price

CRI vs TM-30: which number tells you more about red?

TM-30, and it is not close: it walks 99 colour evaluation samples rather than eight, and reports two numbers where CRI reports one. Bridgelux puts the count plainly, that "CRI uses only eight color samples to determine fidelity, the TM-30-15 uses 99 color samples". The Department of Energy states the underlying problem in one line: "Average fidelity alone is a weak predictor of human perception, especially CIE Ra (i.e., CRI)."

DOE's own experiments show how badly it mispredicts: "Nine of the top 12 rated products had a CIE Ra value of 73 or less, which indicates that the criteria of CIE Ra >= 80 may be excluding many preferred light sources." Nine of the twelve best-liked sources scored Ra 73 or lower, which is the case against buying on Ra by itself. The same research found "gamut shape is more important than average gamut area for modeling human preference, with red playing a more important role than other hues".

DOE describes Rf as an improved version of the CIE colour rendering index and Rg as an improved Gamut Area Index, together giving "two complementary assessments". The tier row comes from a 2019 experiment where 25 participants rated 90 lighting scenes. Rcs,h1 is local chroma shift in the first hue bin, the red one, and its presence in the criteria tells you fidelity alone was not carrying the answer.

Does CRI 98 mean the same thing at 3000K and at 6000K?

No, and this is the least-known thing about the metric. A source is scored against a reference at its own correlated colour temperature, and which reference gets used changes with the number: below 5000K the comparison is a Planckian black body radiator, at 5000K and above it is a daylight illuminant. Two variants of one fitting can both print 98 while sitting different exams.

Houseland lists this spotlight in Warm White 3000K, Neutral White 4000K or Cool White 6000K. Across that spread, 3000K and 4000K would be judged against a black body and 6000K against daylight, so comparing cool white with warm white on one shared number compares two scores that were never on the same scale. That discontinuity at the 5000K crossover is a standing criticism of CIE 13.3, alongside its ageing colour space and the way an arithmetic mean buries a large error in one sample.

Colour temperature also shifts what reaches your eye before rendering enters the argument, and a cool white lands its cast on pale objects first, as our note on white ceramics against black shelves gets into.

How do you check the red rendering in your own room?

Hold a saturated red object under the light, then carry it to a window. Your eye is the better instrument here, provided you hand it a reference to compare against rather than asking it to judge red from memory.

1. Find your stand-in for TCS09: terracotta, a ripe tomato, a red wool jumper. Strong red, not pink or rust.
2. Look at it under the fitting after dark with nothing else switched on. Mixed sources ruin the test, because your eye adapts to the blend and stops reporting the fault.
3. Carry the same object to a north-facing window between 10am and 2pm. Daylight is the reference the metric itself leans on above 5000K, and it costs nothing.
4. Add the back of your hand. Skin rides on that same saturated red, and TCS13 is one of two samples the standard names as a skin tone while leaving it outside Ra.
5. Ask the seller for numbers rather than adjectives: an R9 value, or an Rf and Rg pair. A source somebody measured has them.
6. Where red is the job, as with artwork, food or a make-up mirror, buy on a stated R9 of 50 or better and treat any Ra quoted with no R9 next to it as unanswered.

Red that goes grey in the room and returns at the window is the spectrum, and no dimmer setting or lamp shade recovers it.

What I'd do

Keep the fitting and stop reading the 98. Judge a spotlight on what it demonstrably does — where it lands its beam, what it draws, whether it suits the ceiling it is bolted to — and settle its colour with a red object and a window. On the next light you buy, make R9 the question you lead with. Sellers who have the figure quote it, and the ones who do not tell you something by staying quiet.

One practical note if you are swapping fixtures rather than bulbs: this is a hardwired surface mounted fitting, so it is electrical work, not a decorating job. In England and Wales, electrical work in dwellings falls under Part P of the Building Regulations; in the United States the same job answers to the NEC as your area adopts it, with local permit rules on top. Get a qualified electrician for that step. If you are still choosing, the wall lamps collection is where these sit, and the question worth carrying into it is not what the CRI is, but what the R9 is.

Shop wall lamps from Houseland

-20%Globe Wall Sconce with Switch, Pull Chain Glass Wall Lamp, 4 Shapes 4 Colorsfrom $104.46$130.58-20%Bear Nursery Wall Sconce, Adjustable Iron Wall Lamp with Wood Base, 8 Colorsfrom $97.89$122.36-20%Macaron Bedside Wall Sconce, Adjustable Wood Base Wall Lamp, 3 Shapes 13 Colorsfrom $86.47$108.09-20%Vanity Sconce with Exposed Globe Bulb, Adjustable Iron Wall Lamp, 5 Colorsfrom $72.22$90.28-20%Fluted Resin Hallway Sconce, Up Down Wall Light in 4 Colors, Two G9 Bulbsfrom $83.79$104.74Wood Flush Mount Ceiling Light, Walnut or Natural, Slim or Tall LED Fixturefrom $99.95

See all wall lamps →