Five dimensions we score a cross-check on, and the weight each one carries
A women’s section in a sheet usually holds 2 numbers that look like the same number: the label printed on the tag, and a length in centimetres that somebody typed into a cell. The rubric below scores one thing only — how well a cross-check resolves a disagreement between two such charts — and it stops short of converting labels, because no chart can promise that a tape will land within 1 cm of its row.
| Dimension | Weight | What earns the full weight |
|---|---|---|
| Repeatability | 30 | Measuring the same garment twice lands inside 0.5 cm |
| Stated basis | 25 | The chart says flat-lay or body measurement |
| Named measuring point | 15 | It names the seam the tape runs along |
| Unit and rounding | 20 | Centimetres or inches stated, with rounding shown |
| Recency | 10 | The row carries a date from the current season |
The weights are ours, fixed after walking through 4 chart pairs, and they sum to 100 so a score reads as a percentage. Basis and repeatability carry 55 points between them because 3 of those 4 disagreements came down to one of the two: a flat-lay number read as a body number, or a tape pulled to a different point along the same seam.
This note prints no label-to-centimetre table on purpose. A conversion copied out of a sheet is a claim about a seller, not a measurement of the garment in your hands.
The ranked list, from the method that settles most to the method that settles least
The 5 cross-checks are ranked against the rubric above. The scores are our reading of the 5 methods, not measurements of any garment, and a method can win on repeatability while still losing the argument it was meant to settle.
| Rank | Cross-check | Score | What a pass proves |
|---|---|---|---|
| 1 | Measure a garment you own, twice, against the chart that states its basis | 88 | Whether the chart describes a flat garment or a body |
| 2 | Compare two listings from one seller and look for the same label twice | 71 | Whether the seller types charts or copies one across a catalogue |
| 3 | Ask the seller, through the agent, to measure the actual item | 63 | One item rather than one label, if the reply survives as text |
| 4 | Ask the agent to measure the item in hand at the warehouse | 58 | What arrived, after it has arrived and not before |
| 5 | Read the model note — height and weight — as a sanity check | 34 | Nothing measurable; it flags an obvious mismatch only |
Rank 1 wins because your own tape is the only instrument on the list you can re-run on a second garment an hour later and get the same answer. Rank 5 scores 34 because a model note describes a person and a chart describes a garment: 2 objects, 1 comparison, no conclusion. The [size-compare](/rig/size-compare/) tool exists for exactly that first re-run.
Reading each rank: what it closes, and the case it misses
- Rank 1 — run the tape twice on one garment. Readings more than 0.5 cm apart mean the fabric stretches or the point moved; fix the point before blaming the chart.
- Rank 2 — a seller who copies a chart leaves the same rounding in both listings. Rows that all end in .0 across 3 size steps are a copy signal, not a measurement.
- Rank 3 — ask for 1 measurement in 1 unit, and ask which seam the tape runs between. A reply of "normal size, take your usual" answers nothing.
- Rank 4 — warehouse measurement happens after the item arrives, so it settles the item you have, never the one you were about to order.
- Rank 5 — a model note can rule a size out when the mismatch is large, and can never rule one in.
Two cases show the split. In the first, 2 charts disagreed by a single size step and the flat-lay row matched the buyer’s own garment to within 0.5 cm; rank 1 closed it in about 5 minutes. In the second, both charts carried the same label in the same row and the seller reply named a different measuring point; rank 3 was the only method that could have caught it.
The [paste-link](/steps/paste-link/) stage is where this is cheapest to fix. That stage sits at day 0, and a question sent with the link costs 1 message; the same question sent after the parcel is packed costs a return and 2 freight legs. The [sizing](/table/sizing/) topic page holds the longer version of the same dispute, and the [qc-photos](/steps/qc-photos/) step explains what a photo of a tape can and cannot prove.
Where a ranked list is the wrong instrument
The rubric assumes 2 charts exist. One chart and no second source drops every method to rank 3 or lower, and the honest result is "no comparison available" rather than a low number with a decimal in it.
- Made-to-measure and altered items, where the tag is a starting point and the listed number describes the base pattern
- Knitwear with more than roughly 5% stretch, where a measurement taken at rest understates the worn size
- Bundled sets sold as 1 listing: 1 chart, 2 garments, no way to tell which one the numbers describe
- Anything whose chart was last touched 2 seasons ago, where recency scores 0 and the total cannot pass 90
A ranking is a way to spend 2 minutes, not a guarantee. When the top 2 methods contradict each other, stop and order one item: measure what arrives, note the difference on the chart, and let the second order wait for the tape.
No score here turns a printed size label into a measurement. When a chart and a garment disagree, the garment wins and the chart gets the note.
Worked example: a five-dimension rubric (weights 30, 25, 20, 15 and 10) applied to five cross-check methods, scoring 88, 71, 63, 58 and 34. The weights and scores are this site’s own, and no seller measurement value is reproduced anywhere in the ranking.
This note explains the mechanics; the list is where the items are.