Sizes that mean nothing without a tape measure

Building the two-column table: labelled size against measured millimetres

The table you want has five columns and only two of them require a tape. Column one is the model, column two is the labelled size, columns three and four are the two measured points, and column five is the step between this labelled size and the one below it. That fifth column is the one that turns a list of numbers into a decision.

Build it for three units of the same model rather than one, because a single item tells you about that item and three items tell you about the model. The difference between those two statements is the whole reason this desk retires a model when its three measurements disagree by more than 6 millimetres.

Start with items you already own. A table built from your own wardrobe needs no trust in anybody, costs one evening, and immediately gives you a comparison for every later order. The published tally exists to extend that table, not to replace it.

The measuring basis: insole length, widest point, chest width, body length

Four points cover most of what a size label is trying to say. On footwear, insole length runs from the inside of the heel to the inside of the toe with the foot out of the shoe, and the widest point is taken across the sole at its broadest. On upper-body garments, chest width is measured flat from seam to seam below the armhole and doubled nowhere, while body length runs from the collar seam to the hem.

Each point was chosen because it can be repeated by a second person with the same tool. That is the whole test for a measuring basis: could somebody who dislikes you reproduce your number? If the answer depends on where the tape starts, the definition is not finished yet.

The two runs on this desk followed the same rule. The footwear run published 46 rows and the garment run published 38, three units per model in both cases, using one tape and one measuring board so that a difference between rows is a difference between items rather than between instruments.

Each definition is written out rather than assumed because a tape measurement is only comparable when the start and end points are the same. Two people measuring the chest of the same fleece can differ by three or four centimetres without either of them making a mistake, simply because one started at the armhole seam and the other at the shoulder point. Writing the point down is what turns two numbers into a comparison, and it is also what makes a published row usable by somebody who has never met the person who measured it.

1. Insole length: heel inside to toe inside, foot out, tape flat on the footbed, recorded in millimetres.
2. Widest point: across the sole at the broadest place, recorded in millimetres, taken with the shoe on a flat surface.
3. Chest width: flat, seam to seam, immediately below the armhole, recorded in centimetres without doubling.
4. Body length: collar seam to hem, garment flat and smoothed once, recorded in centimetres.

Reading the table by delta rather than by label

A label tells you what the factory called the item; a delta tells you what happens when you move one label up or down. That is the number your decision actually uses, because the question in front of you is never what a 42 is, it is whether a 42 will behave like the 43 you already own.

On the footwear rows, the step between adjacent labelled sizes runs between 8 and 11 millimetres, which is roughly the width of a thumbnail and about the amount of error a rounded measurement can hide. Read the step before you read the absolute figure, because a model with a 9 millimetre step and a model with a 5 millimetre step are different propositions even when their labels agree.

The delta also explains why a row about one labelled size does not transfer to the next one along. If the step is 9 millimetres and your uncertainty is 6, then moving one label is a decision the table can support and moving two is a guess with a number attached.

1. Small step, small uncertainty: the table supports moving one label in either direction.
2. Large step: measure the item you own against the published points before choosing.
3. No step published: treat the row as a single data point and look for a second model in the same family.

Outliers: the four rows whose labelled size sat two steps away

Four models in the garment run carried a labelled size that ran two steps away from what the measured figures implied, which means the label on the tag and the geometry of the garment disagreed by a whole size on each side. Those rows are the reason a table needs an outlier column and not just a measurement column.

An outlier is not a bad row; it is the most useful row in the table, because it is the one that would have caught you. A model that runs a full size large is exactly the model whose label you would have trusted, and the measured figures are the only thing standing between that trust and a return.

Handle outliers by writing the direction into the row rather than by remembering it. A note that says the labelled size runs large is worth more than a millimetre figure on its own, because it survives being read quickly and it can be quoted in a message when you ask the warehouse to confirm.

Outliers also say something about the label itself. A labelled size is a commercial decision by a factory that wants one pattern to serve several markets, so two steps of drift is not an error; it is a label being cautious or generous in a way that suits a different body from yours. The measured figures are the only part of the row that ignores that decision, which is why they are worth the tape.

1. Flag the direction: runs small or runs large, written next to the measured figures.
2. Check the step: an outlier with a small step is easier to compensate for than one with a large step.
3. Record the run count: a flag built on three units is a property, and a flag built on one is an anecdote.

Folding your own three garments into the same table

Adding your own items takes one evening and follows the same four points. Measure three units of the same model if you have them, or three different garments whose fit you already trust, and write the labelled size beside each measurement. The point of the exercise is not accuracy for its own sake; it is that you end up with a personal calibration against the published rows.

Once the table has six or eight of your own rows, a pattern usually appears: one brand whose labels sit consistently above the measured figure, another that sits below. That pattern is what you are buying when you order a model you cannot measure in advance, and it is worth more than any single published number.

Then keep the table in the same place as your order notes, and add one row every time a parcel arrives. A table that grows by one row per parcel stays current without anybody deciding to maintain it, which is the only maintenance plan that survives a busy month.

1. Measure before the parcel is built, not after it arrives, and ask for the two points rather than a picture.
2. Write the labelled size in the same format the row uses, so the comparison is mechanical rather than interpretive.
3. Add the arrival row within a day of opening the parcel, while the fit impression is still specific.
4. Re-measure one old row each quarter; a tape that has been dropped reads a few millimetres short.

The data point behind this note

Own measurement run, KB-2026-08: 38 upper-body garment rows published, three garments measured per model at flat chest width and body length, with the four models whose labelled size ran two steps from the measured figure flagged rather than removed.

◇KB-2026-08· 2026-09-30 Batch ledger · Challenge a figure

Notes in the same cluster

Related entries and gauges

• What does a size label actually mean on a Chinese listing?
• How do you measure a parcel so the numbers hold up?
• Applicability protocol

  Work out which entries apply to your order right now.