How We Work This Out

The one-sentence version

What we compile

The numerator of every figure here is the manufacturer's own published maximum output. We read it from the manufacturer's product page or manual, on a date we state, and we cite it at the foot of the page it appears on. We do not take specifications from retailer listings, from other publishers, or from memory.

Alongside output we compile whatever else the manufacturer publishes: runtime per mode, beam distance, cell type and capacity, IP rating, impact rating, weight, dimensions, emitter and charge time. Where a figure is absent we write 'not published' rather than estimating it. Those absences are not padding — they are frequently the most informative thing on a specification sheet.

What we compute

Lumens per dollar: published maximum output divided by the live retail price at the moment the page was rendered. Nothing else. It is deliberately the simplest possible arithmetic, because the point is that you can check it.

The price comes from the live price layer with a visible timestamp. Where the layer has no price for a product, the derived figure is left blank and the buy button reads 'Check price'. We never substitute an MSRP, a remembered figure, or a typical street price — a value verdict computed from an invented price would look exactly like a real one, which is precisely why it must not exist.

What we score, and how heavily

There is one further adjustment applied by hand and stated wherever it applies: a light whose headline output has no published runtime beside it is scored below one whose does. Without that adjustment, a ratio built on unverified peak claims would reward the manufacturers who inflate hardest — the exact opposite of what this site is for.

Why battery architecture carries the most weight

Because it is the difference between a purchase and a subscription. A light built around a standard 18650, 21700, 14500 or 16340 can be re-celled for a few dollars indefinitely, so its life is limited by its body and your interest in it. A light whose battery is described as built in or non-removable cannot be re-celled by an ordinary owner: when the cell reaches the end of its cycle life, the body, the optics and the switch are refuse with it.

Almost no manufacturer publishes how many cycles that is. Across every specification we cite, we found two figures: Petzl rates its CORE pack at 2 years or 300 charging cycles, and Maglite quotes up to 2,000 charge and discharge cycles for the LiFePO4 pack in the ML150LRS. That silence is the whole argument for preferring a cell you can replace.

What we refuse to claim

• That we tested, carried, owned, measured or handled any product. We have not.
• Any credential the author does not hold, and any claim of tenure or expertise.
• A price we did not pull live, including an MSRP or a typical street price.
• A rating or review count we did not read from a live source at build time.
• A specification the manufacturer does not publish, including an inferred water rating.
• An ASIN or product identifier we have not verified through a live lookup.

That last one has a visible consequence. Because our Amazon Associates account is not yet eligible for the product data API, we hold no verified product identifiers, so every buy link on this site uses a tagged Amazon search built from the product's name rather than a direct product link. It is a slightly worse experience and it is the honest option: an invented identifier would send you to the wrong item.

What only hands-on testing could tell you

Real sustained output rather than published output. Beam quality, tint and artifacts. How a switch feels after a year. Whether the anodizing survives a toolbox. How consistent two units of the same budget light are. None of that is in any specification and none of it is available to us.

Where a publisher does measure those things with instruments, that work is worth reading alongside this site, and we link to it where it exists. Pointing you at somebody who tests costs us a click and buys the thing that click cannot: a reason to believe every other number on the page.

How and when the numbers change

Prices refresh automatically and the rankings recompute when they do — a value ranking that does not recompute is a value ranking that stopped being true. Published specifications change only when a manufacturer revises one, at which point we re-read the source and update the citation date.

Picks are reviewed quarterly for discontinued models and better-value entrants, guides are reviewed twice a year, and links and product references are checked monthly. Our editorial policy sets out how a pick gets changed and what we do when we get something wrong.