FRPPR بيحسب مقاسك إزاي.How FRPPR predicts your size.

FRPPR gives you one size and a match score — not a table to puzzle over. Two engines power it, depending on what chart is available.

1. Store & standard charts: body estimation

Store widgets use body-measurement charts (chest or bust, waist, hips — in cm). Because shoppers rarely know those numbers, FRPPR estimates them from height, weight and BMI using regressions calibrated against common retail charts, separately for women and men. Body shape (slim, athletic, curvy, fuller) shifts individual measurements by a few centimetres, and waist estimates grow slightly after age 40.

2. Matching and scoring

For every size row, each measurement is compared with the row's range. The distance from the range centre is normalised by the range's half-width d:

• Inside the range (|d| ≤ 1) the penalty is d² — gentle.
• Outside it the penalty is 1 + 4·(|d| − 1) — steep, so sizes that contain you win.
• Chest/bust carries full weight, waist 0.85 and hips 0.8.

Fit preference adds ease before scoring: fitted −2 cm, relaxed +4 cm. If you tell us your usual size, rows away from it get an extra penalty of 1.6 × (steps away)², because a size you actually wear is stronger evidence than a BMI estimate. When the runner-up is an adjacent size with a close score, FRPPR shows it as the “between sizes” alternative.

3. Local-brand charts: the size translator

On the shopper side you photograph a brand's own size chart. An AI model reads it into a table — sizes, measurement ranges, unit, and whether the numbers describe the body or the garment. We keep the table, never the photo. Inches become centimetres, flat (half-width) garment measurements are doubled, and garment measurements lose the usual wearing ease for that category and cut, so every chart is compared in body centimetres.

Your body numbers start from the height-and-weight estimate above, then move toward the sizes you wear in brands you know: 75% your known size, 25% the estimate — or 85/15 when the two disagree by more than 8 cm. The scoring in section 2 then picks the size. Charts that only list height and weight are matched on those directly; there, fit preference moves you to the neighbouring size only when you fall between two.

4. Shoes

Your usual EU size becomes a foot length using the Paris-point relation foot ≈ EU ÷ 1.5 − 1.5 cm, then it is matched to the chart's foot-length ranges. Stores can mark their shoes as running small or large, which shifts the foot length by one size (≈0.66 cm).

5. Confidence

The match score combines how well the best size fits (e^(−score)), how clearly it beats the runner-up, and whether it agrees with your usual size. In the size translator it rises when the chart's measurements come from sizes you know, and drops for garment-measurement charts and height-and-weight-only charts. It is clamped between 51% and 97%. This is an internal ranking heuristic, not a calibrated probability or an independently measured fit-accuracy rate.

6. Selecting the size on a store

The widget recognises dropdowns, radio groups and size buttons, and normalises labels so “Medium”, “M”, “وسط” and “EU 42.0” line up with your recommendation before it selects the size for you.

Limitations

Estimates from height and weight can't see everything — adding your usual size or a store's own measurements improves accuracy. Garments with unusual cuts, stretch or oversized styling may fit differently from their chart.

Prepared by the FRPPR team. Last reviewed: 6 October 2026.