Which keyboard specs actually matter

A keyboard product page contains perhaps thirty specifications. Around five of them predict whether you will like the board. The rest are either irrelevant, unverifiable, or actively misleading.

Here is how to read one.

Specs that genuinely matter

Layout. The single most consequential choice, and the one people get wrong most often. It determines what you can do without holding a modifier. See keyboard sizes explained.

Switch type and weight. Linear, tactile or clicky, and roughly how heavy. This is the difference you feel on every keystroke. See switch types explained.

Hot-swap or soldered. Determines whether the switch decision is reversible. Matters enormously if you are new, very little if you know what you like. See hot-swap explained.

Remapping support. This is the one most people do not think about until it is too late, and it is the reason the Vellum TKL tested here scores badly despite being physically well made — it permits no remapping of any kind. Look for VIA or QMK by name. “Programmable” on a listing means nothing; “customisable software” usually means a Windows-only vendor utility.

Keycap material and profile. PBT rather than ABS if you can, because ABS develops a shine on the home row within weeks. Profile affects both feel and sound.

Specs that matter less than they appear

Polling rate. 1000 Hz means the board reports its state every millisecond; 125 Hz means every eight. For typing this is imperceptible. For fast-paced gaming it is a small real difference.

The reason to check it anyway is diagnostic: a board still shipping 125 Hz in 2026 is telling you something about how much engineering attention it received overall.

Plate material. Steel is stiff and bright, aluminium is stiff and dense, FR4 and polycarbonate are softer, POM is softest. This does affect the sound and feel — but it interacts so heavily with the mount, the case and the foam that the material alone predicts very little. An aluminium plate in a gasket mount and an aluminium plate in a tray mount are different keyboards.

Weight. Heavier boards stay put and usually sound lower-pitched. Beyond that it mostly tells you whether you can carry it.

Battery capacity. Quoted mAh figures are close to meaningless because they are always paired with a runtime claim measured with backlighting off, which is not how anyone uses a backlit keyboard. Assume real-world runtime is a fraction of the claim, and that backlighting roughly quarters it.

Specs you cannot trust

“Gasket mount”. This should be a meaningful specification and it is not. It has been applied to boards with genuine suspended-plate mounting and to boards with a strip of foam somewhere near the plate, and there is no way to distinguish them from a listing.

The only reliable indicators are a flex measurement or a review that describes typing feel specifically. A real gasket mount produces 2 to 3 mm of deflection at the centre; a decorative one produces well under 1 mm. On this site the Castor 65 measured 2.9 mm and the tray-mounted boards measured 0.4 to 1.4 mm.

Switch actuation force. Consistently optimistic. Every switch measured on this site came in above its rating, typically by 1 to 3 g. Treat published figures as comparative rather than exact, and check whether you are reading actuation force or bottom-out force — they can differ by 15 g for the same switch.

Switch lifespan. “80 million keystrokes” is a manufacturer figure nobody independently verifies. Switches rarely fail from wear anyway; they fail from contamination or get replaced by choice.

“Pre-lubed”. True in the sense that lubricant was applied, and it is a genuine improvement over nothing. It says nothing about how much, how evenly, or whether the stabilisers received any — which is the part that actually affects the sound.

Specs that are almost pure marketing

N-key rollover. Nearly every mechanical keyboard sold today has full NKRO over USB. It appears prominently because it sounds technical.

“Anti-ghosting”. The same claim, in older language.

RGB zone counts. A number describing lighting.

“Ergonomic design” on a board with a fixed non-split layout and standard key spacing. It is a rectangle.

What to do instead

The specifications that matter most are the ones least likely to appear on a listing: whether the stabilisers were tuned, how consistent the switches are unit to unit, and whether the case resonates.

None of that is discoverable from a product page. It comes from reviews by people who owned the board long enough for the answers to emerge — which is the entire argument for this kind of site existing, and the reason every review here records what changed at month two and month five rather than only what arrived in the box.

Read the specification list to rule boards out — wrong layout, wrong switch type, no VIA. Then read someone’s account of living with the ones that survive.