Lumens are counted at the emitter and light is judged at the wall

These are listed in the order worth acting on, which with luminous output claims is not the order they are usually presented in.

Three different quantities with similar names

A lumen measures the total visible light a source emits in all directions, weighted for the sensitivity of the human eye. A lux measures how much of that light lands on a surface, which depends on distance and on how tightly the beam is focused.

A candela measures intensity in one particular direction, and it is the quantity that governs how far a beam usefully throws. Two torches with identical lumen totals can differ by an enormous factor in candela, because one floods a room and the other lights a distant object. Choosing between them is a question about the task rather than about which specification sheet carries the larger number.

The thirty-second problem

Small high-output lights run their emitters far above what the body can cool indefinitely, so output falls as temperature rises. The widely used consumer standard for portable lights takes an initial reading shortly after switch-on and also requires a sustained figure, but only the first is usually advertised. A light that starts very bright and steps down within two minutes is not faulty; it is doing exactly what its thermal design requires.

The measurable part is this: sustained output is a far better predictor of usefulness for anything that stays on, such as a head torch on a night walk. The gap between the initial and sustained figures is one of the most informative numbers on a lighting specification sheet, when it is published at all.

Beam shape is the specification nobody prints

A reflector or lens turns the same emitter output into a narrow spot, a wide flood or a mixture, and that choice dominates the experience of using the light. Manufacturers occasionally publish a beam angle or a candela figure, and the two together let you calculate how far the light usefully reaches. Beam artefacts such as rings, dark centres and sharp edges are a design outcome, not a defect, and they matter for close work.

Judged against the category, fixed household fittings have the same problem, since a lamp rated in lumens tells you nothing about how much light lands on the page you are reading. This is the point at which specifications stop helping and photographs of the beam on a wall, published by reviewers, start being the evidence.

Colour is a second axis entirely

Correlated colour temperature describes whether light appears warm or cool, and it is a preference rather than a quality measure. The colour rendering index describes how faithfully colours appear under the light compared with a reference source, and it is genuinely a quality measure. High output and high colour rendering trade against each other in current technology, so a very bright emitter often renders colour poorly.

For anything involving inspection, cooking, first aid or paintwork, the rendering figure predicts satisfaction better than the lumen total does. A specification that omits colour rendering entirely is common at the cheaper end, and the omission is a reasonable thing to notice.

How the numbers get inflated

Emitter manufacturers publish output figures for the bare component under laboratory drive conditions, and some product listings simply reuse those figures. Every lens, reflector and protective window absorbs a share of the light, so out-the-front output is always lower than emitter output. Drive current is often below the emitter's maximum for thermal reasons, which lowers real output further below the quoted component figure.

After the warranty ends, where a listing quotes an implausibly round number with no reference to a test standard, treating it as a component figure is usually the correct reading. The gap is largest in the cheapest products, where there is no certification cost and no reputational penalty for a generous claim.

Manufacturer figures are measured under conditions the manufacturer chose.

Buying by task rather than by number

Decide first whether you need to see a wide area close by or a small area far away, because no single light does both well. For sustained tasks, prioritise the regulated or sustained output figure and the runtime at that level, not the headline burst. For any task involving colour judgement, prioritise rendering and colour temperature over raw output, and accept fewer lumens for the trade.

Check whether runtime is quoted to a stated percentage of initial output, because a runtime measured to near darkness is not a useful number. This site does not measure light output; reviewers with integrating spheres and light meters publish the figures that make these comparisons real.

Everything above, in order of what to do first

1. Three different quantities with similar names. A lumen measures the total visible light a source emits in all directions, weighted for the sensitivity of the human eye.
2. The thirty-second problem. Small high-output lights run their emitters far above what the body can cool indefinitely, so output falls as temperature rises.
3. Beam shape is the specification nobody prints. A reflector or lens turns the same emitter output into a narrow spot, a wide flood or a mixture, and that choice dominates the experience of using the light.
4. Colour is a second axis entirely. Correlated colour temperature describes whether light appears warm or cool, and it is a preference rather than a quality measure.
5. How the numbers get inflated. Emitter manufacturers publish output figures for the bare component under laboratory drive conditions, and some product listings simply reuse those figures.
6. Buying by task rather than by number. Decide first whether you need to see a wide area close by or a small area far away, because no single light does both well.

The takeaway

Ask where the light needs to land, then look for the specification that describes that place rather than the emitter.

The question is rarely which is best. It is which is enough.