Why two size charts for the same item disagree

Build one chest grid from two charts

Two charts for the same item can agree at the smallest label and disagree at the largest without either one being wrong. The reason is grid arithmetic rather than measurement error. Take a chart that steps 6 cm of chest per label from 96 cm at the smallest size, and a second chart for the same item that steps 4 cm per label from the same 96 cm. Both are internally consistent, and they simply do not stay aligned with each other.

Six rows and three columns make up the delta matrix. Row labels run S, M, L, XL, 2XL and 3XL. Column A is the 6 cm grid: 96, 102, 108, 114, 120, 126. Column B is the 4 cm grid: 96, 100, 104, 108, 112, 116. The third column is A minus B, giving 0, 2, 4, 6, 8, 10, and every entry is arithmetic on the two grids stated in the same unit the charts use.

Read the diagonal and the pattern fits in one sentence: drift equals the step difference multiplied by the number of labels above the base. Here the step difference is 2 cm, so the gap opens 2 cm per label and reaches 10 cm at the fifth label above the base. A buyer comparing the two charts at the largest size faces a question five times bigger than the one they think they are asking, and at the smallest size faces no question at all.

This grid is a tool rather than a measurement. It says which chart produces the larger garment at a given label, and by how much, given the step each chart uses. It cannot say which chart is accurate for a particular item, because neither chart is a tape measure. Use it to decide which question to ask, whether about the step or about the base, and then ask that question of the seller rather than of the table.

Record where each number in the grid came from

Both grids start from a base of 96 cm, a placeholder chosen so the arithmetic divides cleanly and the matrix stays legible. The steps of 6 cm and 4 cm are the two grading conventions being compared here, not readings taken from any listing. Everything in the table is derived from four inputs: two bases that coincide at 96 cm, and two step sizes.

Keep a statement separate from a measurement. A published chart is a statement a seller makes about a garment; a tape applied to the garment is a measurement; a matrix like this one is arithmetic performed on statements. Where a statement and a measurement disagree, the measurement wins, and no amount of chart comparison substitutes for one tape. Hold the three categories apart in your notes, because they fail in different ways.

Our own listing sample gives the reason to care about multi-style items at all. In the 2026-10-06 snapshot of 192 titles, 26 carry the multi-variant flag, and 24 of them state a plain style count running from 2 to 40 with a median of 14.5. A title declaring 40 styles is a title that can attach forty different measurement grids to a single order line, and the two charts in front of you may not even belong to the same grid.

Reading the same file twice produced two different counts, which is worth recording rather than smoothing over. A strict reading of the square-bracket form finds 24 titles, while the flag in the data file marks 26, because two of them declare something other than a plain style count: one writes 40 plus, one counts colours. A looser scan that also accepts the parenthetical form finds 11 further titles. Strict and loose readings of one unchanged file differ by 11 items.

Check the annotations attached to each column

Start with a note on fields: identify the field each column names before comparing values. A column headed chest on a garment-measured chart and a column headed chest on a body-measured chart are two different quantities wearing the same word, and subtracting one from the other produces a number with no meaning. The matrix above is valid only because both columns use the same field and differ only in step.

Second note: find the tolerance line. A chart printing a tolerance of plus or minus 2 cm is telling you that a single-row difference of 2 cm sits inside the noise. That same tolerance does not excuse a 10 cm difference spread across five labels, because rounding noise does not accumulate along a constant step. One row of difference is ignorable; five rows of difference is a decision.

Third note: locate the measurement point. Armpit to armpit, two centimetres below the armhole, and across the fullest part are three different instructions that move the number while no drift is present at all. A chart that never states its measurement point is a chart whose numbers cannot be compared with anything, including a second chart from the same seller for the same item.

Fourth note: record the fabric. Knitted goods stretch and woven goods do not, so an identical stated chest fits differently on a sweatshirt and on a shirt. This note changes no cell in the matrix, and it changes which row you should be reading, which is why it belongs beside the table rather than inside it.

One more note matters most of all: never compare a single row. A one-row comparison cannot separate a step difference from a rounding difference, and rounding is applied per cell. The matrix has six rows because one row is ambiguous and six rows are not. If all you hold is one row of one chart, you have a label rather than a comparison.

Explain the zero row and the widest row

Counterintuitive row one is the smallest size, where the drift is zero. Zero reads as agreement, and the natural conclusion is that the two charts match, which is precisely the wrong lesson, because the base is the one place the two grids were defined to touch. A zero in the base row says nothing about any other row, and it is the entry most often used to justify ignoring the rest of the table.

At the other end sits the largest row, where the drift exceeds the tolerance printed on the chart. Ten centimetres against a stated plus or minus two looks like a misprint, and it is not, because it is the predictable output of a 2 cm step difference applied five times. Values that grow smoothly along a row are arithmetic, and values that jump are candidates for a misprint. Smoothness rather than size is the test to apply.

Separate the two diagnoses by checking each column on its own. In an arithmetic drift both columns step evenly: 96 to 126 in sixes and 96 to 116 in fours, with every gap identical. In a genuine misprint one column breaks its own sequence while the other stays even, so a run reading 96, 102, 121, 114 shows the break in a single cell rather than in the difference between the two columns.

One practical rule falls out of both rows. If you know your own body measurement, read the chart whose header names a body field; if you have the garment in your hand, read the chart whose header names a flat or garment field. Comparing the two charts against each other is the third-best option, and it is the one that generates the largest numbers in this article. Compare like with like first, then measure.