How to measure a dress

Four things: a firm flat surface, a fabric tape or a steel rule, the dress with every closure fastened, and about five minutes. A bed is too soft and carpet is too soft, because both let the fabric sink and spread. A dining table, a clean hard floor or a cutting mat holds the garment still while you read the tape.

Lay the dress face up. Close the zip, fasten the buttons, do up the hook and eye. Smooth it outwards from the centre with your palm rather than pulling on the edges. You want the shape the garment holds on its own, because that is the shape the maker measured when it wrote its size chart.

Everything you take this way is a flat measurement: the width across one face of the garment. A body circumference is roughly double that figure. The distinction matters, because dress charts mix the two conventions and do not always label which is in use. If a chart lists a bust of 34 and your flat reading is 17, you are reading a body chart. Check the heading before you conclude anything.

Definitions for the terms in that table, including high point shoulder, sweep and set-in sleeve, are collected in the glossary.

A t-shirt is close to a rectangle with sleeves. It has two side seams, one hem and very little shaping, so it settles the same way every time you put it on the table. A dress fights back on three fronts.

Stretch. A knit dress with elastane can read several inches wider under light tension than it does relaxed, and there is no agreed amount of tension to apply. Always measure relaxed, and write the word relaxed next to the number.

Drape. A soft rayon or a heavy jersey does not hold a shape on a table. It creeps outwards as it settles, so a reading taken immediately after you lay it down differs from one taken a minute later. Smooth it, wait, then measure.

Bias cut. A bias-cut panel is woven at forty-five degrees to the grain, which makes it hang beautifully on a body and behave badly on a table. It elongates under its own weight and narrows as it does so. Length and width readings on a bias dress trade against each other, and neither is stable to the same degree as a straight-grain seam.

On a woven cotton shift, two careful attempts should land on the same figure. On a bias-cut satin or a ribbed knit they will not, and no technique fixes that. Take three readings and record the range rather than a single number.

Two dresses from the same production run are not identical. Cutting, sewing and finishing all introduce variation, and a garment-dyed piece shrinks unevenly in the dye bath on top of that. ShirtMath does not hold published dress specifications, so the honest comparison comes from the blank t-shirt specifications it does hold, where manufacturing tolerance on a published figure runs about plus or minus 0.3" on a stable knit and up to plus or minus 0.8" on a garment-dyed style. A dress in a soft or dyed fabric will not do better than that, and a bias cut will do worse.

The practical consequence: a difference of half an inch between your reading and a published figure is noise. A difference of two inches is a real difference between garments.

Less than most shoppers assume. A size letter is a name a brand chooses, not a measurement it is obliged to honour. The clearest published evidence sits in blank t-shirt specifications, where the same letter names visibly different garments depending on the cut.

The ladies cut publishes a 17.5" Medium against 20.0" to 20.5" for the unisex Mediums, a gap of three inches on the same letter, and the same three inch gap holds at Large. Womenswear sizing carries that problem further, because numeric dress sizes have drifted over decades and drift differently between brands. That drift is the subject of the vanity sizing guide, and it is the reason a measured garment beats a remembered size.

Once you have your numbers, compare them against a garment chart rather than a body chart. The dress size chart sets out how the common conventions line up, and the women's clothing size chart covers the wider range. Match on the measurement you actually took, in the units and convention the chart states, and ignore the letter on the label.

Nothing, and that limit is worth stating plainly. The ShirtMath scanner is live for one thing: a shirt laid flat with a scale card in frame. It returns flat measurements for that shirt and matches them against published garment measurements across the catalogue. Roughly half of its readings land within an inch of a tape, sleeve opening and bicep are the least reliable points, and anything it cannot read confidently is left blank rather than guessed.

Dresses are not in that scope. There is no dress scan, no body scan and no fit prediction from a photograph of a dress. For a dress, the tape in your hand is the instrument, and this page is the method. How the scanner's accuracy is assessed is written up in the methodology.

The same discipline transfers between garments. The shirt measuring guide covers the equivalent points on a top, and the full set of walkthroughs sits on the guides hub.