There is more than one system, and they disagree
The Female Figure Identification Technique, or FFIT, developed at North Carolina State University, defines nine categories. Commercial women's systems typically use five, and commercial men's systems use a different five again. There is no governing taxonomy.
The systems have not been validated
This is the important part. Researchers publishing in Ergonomics in 2021 analysed 1,679 3D body scans and evaluated FFIT, which ergonomics researchers had been treating as a gold standard. They concluded the system has never been scientifically verified, and that researchers were using it without a scientific basis.
Different but equally permitted interpretations of where to place the tape measure classified the same body into different shapes in up to 40% of cases. Moving the tape by 1 cm was enough to change the category for a large share of bodies.
If a 1 cm shift reclassifies 40% of bodies, the category is not a property of your body. It is a property of where someone placed the tape.
Machine learning does not rescue it
On a public five-class body shape dataset, a VGG16 model only reliably separated the hourglass and apple categories. VGG19 failed to discriminate between types, and ResNet18 confused inverted triangle with hourglass.
So why does anyone use it?
Because it is useful shorthand for styling. “Inverted triangle” tells a shopper which necklines to look at. That is a legitimate use. It is not a measurement, and it should not drive sizing decisions.
What is actually measurable
Shoulder-to-shoulder width, waist-to-hip ratio, and the ratio of bust to underbust are all directly measurable numbers with real meaning. If you want to know how clothes will look on you, those beat a shape label.
Waist-to-hip ratio is the closest thing here with clinical significance, though it is also the least reliably measured: inter-observer correlation in one study was only 0.51.
The takeaway
If a tool tells you you are an hourglass, fine, use it to browse necklines. If it tells you that because of your shape you need a medium, disregard it. The second claim is not supported by any validated research.
Read which measurements actually matter instead.