The same product, three different whites
Pull a shelf of LED pillar candles from a single retail display and hold them up together. One burns a deep amber. One runs closer to cream. One, the one the buyer called “warm white” on the purchase order, looks almost grey next to the other two. All three boxes printed the same marketing line: *flame-realistic, warm glow, battery powered.*
If you are buying these in case quantities, that discrepancy is your problem, not the factory’s. Color temperature is the first line on an LED candle spec sheet that a buyer can actually enforce. And most buyers never ask for it, because the term feels like a physics lesson instead of a purchasing lever.
It is a purchasing lever. Here is how to read it.
Kelvin is the number, not the adjective
Correlated color temperature — CCT, measured in Kelvin — describes the perceived hue of a light source by comparing it to a theoretical black-body radiator at a given temperature. For a real candle flame the value sits somewhere between 1500 and 2000K. For LED white light, manufacturers cluster their chips in a few practical bands:
**1800K–2200K** — “flame white.” Closest to a real wick. Looks best in a dim room, flattens fast under any ambient light.
**2700K** — the industry default for “warm white.” Slightly yellower than a real flame but reads natural to the eye in most retail and home lighting.
**3000K–3500K** — “neutral white.” Fine for a hallway or an event where brightness matters more than realism. Noticeably cooler next to 2700K.
**4000K and up** — “cool white.” Almost nobody wants this on a candle. It is the white of an office. On a candle it reads as an error.
Two practical consequences. First, a 2700K and a 3000K LED placed side by side do not look like two versions of the same candle; they look like two different products, and shoppers can tell within a second. If you are merchandising a set of four or six pillars, a spread wider than roughly 200K between units is visible. Second, “warm” on a supplier’s data sheet is not a spec. “2700K ±150K” is. If the spec line only says *warm white* or *candle flame*, the actual chips could land anywhere from 2500 to 3200K and still ship compliant with what was written.
The fix is one line on your inquiry: **state the CCT, the tolerance band, and the measurement condition** (room temperature, fully charged battery, steady-on mode). Suppliers with a real test bench will give you a number and a bin. Suppliers without one will give you an adjective.
CRI: the number that decides whether the flame looks like wax
CCT tells you the hue. Color rendering index — CRI, rated from 0 to 100 — tells you how faithfully that light reproduces the colors it falls on, relative to a natural reference source. On an LED candle that matters more than it does on, say, a desk lamp, because the whole selling point is that the flame and the wax look like wax and flame.
A cheap 2700K chip with a CRI in the mid-70s will still print as “warm white” on any colorimeter. What it actually does is muddle the reds in the simulated flame and pull the wax toward a flat beige. The candle stops looking like a candle and starts looking like a light. Raise the chip to CRI 80 and the flame recovers a believable gradient. At CRI 85–90 — common in anything positioned as premium — the flame tip, the body, and the melt pool read as separate colors the way a real wick does.
Two buyer notes. CRI 80 is a defensible floor for bulk purchase; below that you are buying a light fixture with a wax shell. And ask for the Ra value measured on the assembled unit, not the bare LED die. The phosphor blend, the lens, and the diffuser on a candle head all shift the final spectrum, so a die that measures 88 in the bin can land closer to 80 once it is inside the flame sleeve. The assembled-unit number is the one that reaches the shelf.
Flicker is a frequency, not a feature
Every “realistic flame” LED candle is a flicker engine: the driver modulates current to make the light dance. The difference between a convincing flicker and one that makes a room feel slightly wrong is frequency.
Slow, low-frequency flicker — a few cycles per second, common on the cheapest drivers — mimics a flame badly. The eye registers it as stuttering, and video capture makes it far worse: at typical 30fps camera shutter speeds, low-frequency PWM flicker becomes banding and moiré, the classic “shimmering artifact” in photos and clips. Events and stores increasingly film their displays, so a candle that looks fine standing still can look broken on camera.
The spec line that matters is flicker frequency and depth. A driver running its modulation above 50Hz with a moderate depth change per cycle reads as a calm flame to the eye and stays clean under most video. What you should ask for is simple: the flicker mode’s refresh rate, and confirmation that the flicker pattern is not a plain 50/50 square wave. Real flame motion is irregular; a factory that can tell you the pattern is random-walk versus fixed-sequence is a factory that can also fix it when a customer complains.
The spec card worth asking for
Strip the marketing and an LED candle’s light is defined by four numbers. A complete spec line for a sourcing file looks like this:
**CCT:** 2700K ±150K, measured steady-on, charged battery
**CRI (Ra):** ≥80 assembled unit (≥85 for premium tiers)
**Flicker mode:** modulation above 50Hz, random-pattern, depth specified
**Brightness range:** lumens or candela, full-on and in the dimmable mode if offered
Anything the supplier cannot give you as a number, treat as a vibe. The first three lines are testable on a bench with a colorimeter and a frequency analyzer; the fourth is a reading off a lux meter at one meter. If a vendor pushes back on any of them, the pushback is itself data — it tells you where their control actually ends and the next bin’s surprise begins.
Buyers who write these four lines into their POs stop shipping a “warm white” that arrives as three different whites. That is the whole trick: the candle industry sells adjectives, and the spec sheet is where the adjectives get converted into numbers.