Measure sit-bone width at home: repeat the impression, not just the ruler
Make independent centre-to-centre observations, keep unclear marks unresolved and distinguish a posture-labelled starting range from a verified saddle fit.
Published 9 October 2026 · OpenBikeFit
Saddle reference field guide · not to scale
New impression. Independent reading.
Fictional readings: 120 mm and 124 mm
Difference 4 mm · mean 122 mm
Agreement does not establish anatomical accuracy or an ideal saddle.
On this page
- 01Prepare one method you can repeat
- 02Identify the centres before measuring
- 03Make a second independent impression
- 04Understand the assistant’s agreement check
- 05Agreement is not absolute accuracy
- 06Keep the posture allowance labelled
- 07Do not infer this input from a body label
- 08Choose a next step that preserves the limits
Name the reference first
Check independent impression readings
The existing free tool keeps its own access and hardware boundaries. Reading a guide or following a link does not approve work or create a bike.
Open toolA sit-bone measurement is most useful when another observation can challenge it. Measuring the same two pencil marks several times mainly checks the ruler reading. A fresh impression also checks how you sat, how the material recorded the contact and where you identified each centre. Those are different sources of variation.
This guide explains a paper-first method for an adult without an active health or injury concern. It does not diagnose symptoms, establish a clinical measurement or tell you which saddle to buy. You can read the whole procedure without entering body information, creating a bike, enabling a sensor or accepting processing consent.
Prepare one method you can repeat
Use a stable seat and a firm, flat support for the measuring material. A suitable corrugated cardboard impression method is a low-tech option, not a calibrated instrument. Follow the instructions accompanying the particular kit if you use one. Keep the support, material, clothing and stated sitting condition consistent between observations.
SQlab publishes a home measurement route and a trained-retailer alternative. That is evidence that the manufacturer uses this input in its own sizing system, not independent proof that every home observation is accurate. If sitting or making the impression cannot be done comfortably and securely, do not force the test. Leave the input unresolved and obtain appropriate assistance.
Identify the centres before measuring
Stand up and inspect whether two impressions can actually be distinguished. Mark each estimated centre separately. A broad patch, a clothing seam, a rivet or an ambiguous group of marks should not become a precise anatomical endpoint just because a ruler is available. Keep uncertainty visible rather than choosing whichever marks produce a familiar saddle size.
Read the straight centre-to-centre separation in millimetres. Do not measure the total outside span of both patches, a curved path around them or the width of the cardboard. If the left centre is at 30 mm and the right centre at 152 mm on one ruler, the separation is 152 − 30 = 122 mm. These are fictional ruler coordinates, not suggested body dimensions.
Make a second independent impression
Use a fresh area or a new piece and repeat the complete process. Do not trace the first centres onto the second impression, or shift the second marks until the readings agree. Preserve both raw distances with a brief method note. A tidy mean cannot reveal whether one of its inputs came from a different sitting condition.
Imagine independent readings of 120 and 124 mm. Their difference is 4 mm and their mean is 122 mm. Both values remain part of the record. Reading 122 mm twice from the same original pencil marks would not provide the same kind of evidence, even though the displayed pair looks more consistent.
Understand the assistant’s agreement check
The existing free width assistant accepts two valid readings within 5 mm. A larger mismatch requests a fresh third observation. Its triplet check keeps every reading against the median; it does not silently select the closest pair. These rules are explicit engineering choices in OpenBikeFit, not manufacturer tolerances or universal anatomical constants.
For example, 120 and 128 mm disagree by 8 mm. If a fresh third reading is 132 mm, the median is 128 mm and the first reading is still 8 mm away. That set remains unreliable under the product rule. Averaging it, deleting 120 or declaring 128 and 132 the only useful pair would hide the unresolved method problem.
Agreement is not absolute accuracy
All impressions can share an offset. The material may spread marks similarly each time, or the same centre-identification habit may affect every reading. A small observed difference therefore does not establish that the true spacing lies within that difference, and it is not a confidence interval with a stated probability.
Keep instrument resolution separate from the variation you observed. A ruler with millimetre divisions does not make the whole process accurate to one millimetre. Adding decimal places to a mean records arithmetic, not additional anatomical knowledge. Where marks are unclear, repeating the method is more informative than adding more digits.
Keep the posture allowance labelled
The tool combines the measured mean with a selected SQlab-derived posture allowance. That allowance is not a measured part of your pelvis. For the fictional 122 mm mean and the tool’s sport/endurance allowance of 20 mm, its centre estimate is 142 mm. The displayed 137–147 mm starting range includes a product uncertainty floor, not a clinically validated prescription.
Other manufacturers can use different sizing systems or discrete sizes. Specialized also presents measurement-led saddle selection, but that does not make every brand’s width label equivalent. Do not automatically round 142 to a catalogue size and call the selection verified. Retain the method, chosen convention and actual candidate’s width definition.
Do not infer this input from a body label
Height, weight, a clothing size or a gender category does not provide the two impression centres required by this workflow. The assistant deliberately does not infer spacing from those fields. A label can describe a product range without becoming a substitute for the measurement you are trying to make.
Likewise, the current saddle’s external width cannot supply a missing body observation. It can be entered separately for a numerical comparison, but that comparison does not measure how the surface supports you. Padding, curvature and the longitudinal place you use remain outside the impression arithmetic.
Choose a next step that preserves the limits
Use the free assistant when you have independent, readable observations and want to inspect its named convention. Use paper if you are still resolving the method. The guide on support area explains why the resulting number cannot be compared indiscriminately with every saddle’s maximum width.
A later real saddle trial needs the exact saddle and seatpost instructions and a physical check of the assembly. This guide makes no mechanical change and recommends no particular product. Do not send completed body-measurement notes through a business inquiry; that inquiry is for separately scoped workflow services, not individual rider records.
Practical questions
Frequently asked questions
How do I measure sit-bone width with cardboard?
Use a suitable impression method on a stable support, identify two readable centres and measure their straight separation. Make a fresh independent impression rather than rereading the same marks. Ambiguous marks stay unresolved.
Should I measure the outside edges of the impressions?
No. This workflow uses centre-to-centre separation. The width of a patch or the complete outside span is a different quantity.
What if the two readings disagree?
Preserve both and repeat the complete method. OpenBikeFit requests a third reading after more than 5 mm disagreement; that threshold is a product rule, not a universal accuracy specification.
Does sit-bone spacing tell me the exact saddle to buy?
No. A named posture convention can produce a starting range. Usable surface, the manufacturer’s width definition and the physical assembly still need separate checks.