How we measure wheels

Two things, because they are the two a buyer cannot check for themselves before buying and cannot easily correct for afterwards:

• Weight — each wheel separately, and the pair as the sum. Claimed pair weights are the single most commonly optimistic figure in the category.
• Rim geometry — internal bead-seat width, maximum external width and rim depth. Internal width decides which tyres fit properly and how wide they actually sit, and it is occasionally published rounded or simply wrong.

We do not claim aerodynamic figures. Meaningful drag numbers require a wind tunnel or a properly controlled field protocol, neither of which we have. The site models an aero orientation from geometry and says explicitly that it is modelled, not measured. We do not test stiffness, durability or braking either — those need rigs and sample sizes we do not have, and a single-sample impression of them would be worth less than nothing.

A reading is only as good as the thing that produced it, so the instrument specification is a requirement rather than a preference.

Scale resolution

1 g or finer

Scale capacity

5 kg minimum

Caliper resolution

0.02 mm or finer

Calibration check

Every session, before use

Digital, 1 g resolution or finer, minimum 5 kg capacity, checked against a certified reference mass at the start of every session. Digital caliper, 0.02 mm resolution or finer, zeroed closed before every reading.

The scale is checked against a certified reference mass before every session and again at the end. If the closing check has drifted, every reading taken in that session is discarded — not corrected. A drift we can only detect at the end of a session is a drift we cannot place in time, and adjusting figures to compensate would be inventing precision we do not have.

The instrument register is empty. When equipment is commissioned, each item appears here by make, model, resolution and date of its last calibration check — and no measurement can be recorded against an instrument that is not listed.

This is where most published wheel weights go wrong, and it is not usually dishonesty — it is two people weighing different objects and comparing the results.

A manufacturer's claimed weight is normally a bare weight: the built wheel with nothing added. Fit rim tape, tubeless valves and rotors and the same pair reads 150–250 g heavier. Add tyres and sealant and it is heavier again by the better part of a kilogram. A review that weighs a shop-prepared wheel and reports the difference against a bare claim has measured its own setup.

So we record every removable component explicitly, present or not:

Rim tape

Present / absent, and which

Valves

Present / absent, and which

Tyres

Present / absent, make and size

Sealant

Present / absent, volume in ml

Cassette

Present / absent, and which

Rotors

Present / absent, and which

Freehub body

Which one — worth 20–60 g

End caps / axle standard

Where removable

The comparison against a claimed weight is only drawn when the wheels were weighed bare. If they were not, we publish our figure with the setup stated and no difference is calculated at all. The code enforces that rather than relying on anyone remembering.

1. Check the scale against the reference mass. Record the reading.
2. Strip the wheel to the state being recorded. Note what was removed and what remains.
3. Zero the scale with nothing on it, on a level, solid surface — not a carpet, not a workbench that flexes.
4. Weigh the front wheel. Remove it fully from the scale, allow the display to return to zero, and repeat — at least 3 times in total. Placing and replacing the wheel is the point: a wheel left sitting on the platform will reproduce the same reading whether or not it is correct.
5. Repeat for the rear wheel.
6. Re-check the scale against the reference mass. If it has drifted, discard the session.
7. Photograph the readouts. Record ambient temperature — load cells drift with it.

Rim dimensions are taken with the tape removed, at a minimum of 3 points spaced around the rim, with the caliper zeroed closed before each reading. A carbon rim is not identical all the way round; one reading at one point is an anecdote.

Every published figure is a mean of repeats, and the spread between those repeats is published alongside it. That spread is our own uncertainty, and hiding it would make our numbers look more precise than they are.

Weight repeats, per wheel

3 minimum

Max spread across repeats

2 g

Dimension readings, per measurement

3 minimum

Max spread across readings

0.15 mm

Treated as agreeing with a claim

Within 15 g / 0.3 mm

A session that breaches a spread limit is rejected and re-run. It is not published with a note explaining that the readings were a bit scattered — the whole value of measuring something ourselves is that it met the standard.

The agreement tolerance is deliberately generous, and it is not there to protect manufacturers. Carbon layup and resin content genuinely vary between samples of the same wheel, and our instruments have uncertainty of their own. Calling a 10 g difference a discrepancy would be a claim about the maker that our method cannot support from a single sample.

One sample is one sample. Who supplied it changes how much that sample is worth, so it is published on every measurement:

• Bought at retail

  Ordered anonymously, so the sample cannot have been selected for us.

• Loaned by the brand

  Press samples can be selected, and we have no way to rule that out — weigh the figure accordingly.

• Loaned by a retailer

  Ordinary shop stock, but not ordered anonymously.

• Reader’s own wheels

  One sample of unknown history and mileage, measured by us.

No brand or retailer sees a measurement before it is published, and none has ever been given the opportunity to comment on one, dispute one, or have one withheld. If that ever changes, this paragraph changes with it.

Where an outlet states it weighed a wheelset itself, we record the figure, attribute it, link the article and date both their publication and our reading of it. Those appear on a wheelset's sheet under Weighed elsewhere, kept visually and structurally apart from anything we measured. They never count towards our own coverage and never change a wheelset's verification status.

The bar for recording one is that the outlet says they weighed it. A review that repeats the maker's claimed figure is not a measurement, however confidently it is written — and in practice most of them do exactly that. Of four reviews of one wheelset we checked, three turned out to be quoting the manufacturer. Those rejections are recorded in the source file so nobody re-checks them.

Where two or more outlets have weighed the same wheelset we publish the range between them rather than an average. The gap is the useful part: it is sample variation, different scales and different setups combined, and it tells a reader something no single measurement — ours included — ever can.

Where we have measured a wheelset, its sheet shows the manufacturer's claimed figure and ours side by side, with the difference stated. We do not overwrite their number with ours: their claim is a fact about what they say, ours is a fact about one sample, and both are worth knowing.

Where we have not measured a wheelset — currently 47 of 47 — the sheet shows the manufacturer's figure, labelled claimed, with its source and the date we checked it. There is no measured column, because there is no measurement.

A measurement here is never a verdict on quality. A wheel 30 g over its claim is not a bad wheel, and one that comes in under is not thereby a good one. We publish the figure and let it be a figure.

Corrections: if you have measured a wheelset on calibrated equipment and your figures differ from ours, we want to know. Two evidenced measurements that disagree are recorded as a conflict and both are shown — the site has carried that mechanism since it was built, and it exists precisely so we never have to quietly pick a winner.