How we normalize manufacturer sizing
A medium is not a medium. One brand's M starts at 168 cm, another starts at 173 cm, and a third does not use letters at all. Putting those charts side by side is most of what BikeSize does, so this page explains what we change on the way in, what we leave exactly as the brand wrote it, and what we will not guess.
The one rule everything else follows
The manufacturer's own chart wins. Our sizing formulas are useful for narrowing a shortlist, but when a brand publishes a fit range for a specific frame, that range is the better answer for that frame. Normalization is about making charts comparable, not about correcting them.
What we normalize, field by field
| What we store | Rule |
|---|---|
| Size label | Kept character for character as the brand publishes it. A Trek 54 stays 54, a Giant ML stays ML. We never translate one brand’s label into another brand’s scale. |
| Category | Each chart is filed under road, mountain, hybrid, gravel, ebike, kids, or cargo, because most brands publish a separate chart per family and merging them would invent a size that does not exist. |
| Rider height range | Stored as a minimum and maximum in centimetres. Imperial charts are converted on the way in, so a chart is never half inches and half centimetres. |
| Inseam range | Also a min and max in centimetres. Where a brand publishes only height, the inseam range comes from the same published row and nowhere else. |
| Per-size geometry | Stack and reach in millimetres, with the official chart URL, the capture method, a last-reviewed date, and a confidence level recorded alongside. |
- Size label
- Rule
- Kept character for character as the brand publishes it. A Trek 54 stays 54, a Giant ML stays ML. We never translate one brand’s label into another brand’s scale.
- Category
- Rule
- Each chart is filed under road, mountain, hybrid, gravel, ebike, kids, or cargo, because most brands publish a separate chart per family and merging them would invent a size that does not exist.
- Rider height range
- Rule
- Stored as a minimum and maximum in centimetres. Imperial charts are converted on the way in, so a chart is never half inches and half centimetres.
- Inseam range
- Rule
- Also a min and max in centimetres. Where a brand publishes only height, the inseam range comes from the same published row and nowhere else.
- Per-size geometry
- Rule
- Stack and reach in millimetres, with the official chart URL, the capture method, a last-reviewed date, and a confidence level recorded alongside.
Overlapping ranges stay overlapping
Brands routinely publish ranges that overlap. Trek's road chart puts a 52 at 163 to 170 cm and a 54 at 168 to 175 cm, so a rider at 169 cm is genuinely in both. We keep that overlap. Trimming it into tidy non-overlapping bands would look neater and would be a lie: the brand is telling you that either frame can be made to fit, and which one is better depends on reach, flexibility, and how you want to ride.
So when you land in two sizes at once, the page says so rather than picking for you. The tiebreakers we point at are the ones a shop would use: the smaller frame for a more upright and manoeuvrable fit, the larger for more reach and stability, and a test ride ahead of either.
What we never infer
- Missing sizes. If a brand does not publish a fit range for a size, we do not interpolate one from the sizes on either side.
- Missing geometry. Stack and reach are copied from the official chart or left out. A blank cell is more honest than a plausible number.
- Cross-brand equivalence. We will show you a Specialized chart next to a Trek chart. We will not tell you that a Specialized 54 equals a Trek 54, because the two are measured against different geometry.
- Discontinued or unpublished model years. Where a brand has pulled a chart, the page keeps what was published and dated rather than substituting the current year's numbers.
How errors get caught
Charts are checked automatically before they ship. Validation rejects impossible measurements, unit mix-ups such as 56 entered where 560 mm belongs, empty or duplicated size labels, a source URL that is missing or not a real link, and a review date in the future. It also rejects geometry that goes backwards: stack and reach have to increase as sizes go up, which is how transcription slips get caught.
Numbers marked as high confidence must cite an official manufacturer source. Partial or inferred data is marked low confidence and is not allowed to carry a public page on its own. The provenance rules are set out in full on how we source and update data.
What we do not do
- We do not measure frames ourselves. Nothing on this site comes from hands-on measurement, ride testing, or lab work.
- We do not adjust a brand's published range to match our own formulas, or the other way round.
- We do not let commission decide which brands get a size chart. See the affiliate disclosure for how links are marked and paid.
Corrections
If a chart here does not match the brand's current published chart, report it on the public issue tracker, ideally with a link to the manufacturer page. Reports are handled in the open. We aim to respond within five working days, and a confirmed error is fixed in the published chart rather than left to the next review.
Charts are re-checked against the manufacturer when a model year changes, and at minimum every twelve months, on the review cycle described in the editorial policy.
Back to all methodology pages.