How JOTT used AI to improve fit accuracy & reduce returns

Founded in 2010 in Marseille by Mathieu and Nicolas Gourdikian, JOTT, Just Over The Top, built its name on a deceptively simple object: the ultra-light down jacket, packable into its own pouch and designed to live in a handbag as easily as on a shoulder. What began as a Marseillais essential has become an international brand with boutiques and corners across Europe and beyond, its colour-block puffers recognisable at fifty paces. The range has grown around the same idea of versatile comfort, vests, parkas, rainwear and knits engineered to layer through every season, with a stated commitment to higher-quality, more responsible materials as the brand scales. JOTT's proposition is ease: one garment, many uses, no fuss. Online, that very versatility became its sizing paradox, a jacket designed to be worn ten different ways is a jacket ten times harder to size from a chart.

Down is a fit problem disguised as a simple one. An ultra-light JOTT puffer might be worn over a T-shirt in October, over a heavy knit in January, or under a shell on the slopes, and the right size is different in each scenario. Loft, sleeve length and shoulder fit interact with layering intent in ways a flat size chart cannot express, and JOTT's cross-seasonal, unisex-leaning range multiplied the ambiguity. The consequences showed up where they always do: shoppers hesitating at the size selector, conversion leaking at the exact moment of decision, and a return rate inflated by customers ordering two sizes to try at home. For a brand whose entire promise is effortlessness, making customers do sizing homework was off-brand as well as expensive. JOTT needed guidance that could ask, in effect, how the jacket would be worn, and answer with a single confident size.

JOTT partnered with Kleep to put AI Smart Sizing on every product page. The flow is as light as the product: from two photos taken on a phone, or a brief questionnaire for shoppers who prefer it, Kleep builds a precise measurement profile and returns a personalised recommendation in under 250 milliseconds. Under the surface, each JOTT garment is calibrated individually, an ultra-light packable and a winter parka are treated as the different sizing problems they are, with the intended fit and layering logic of each cut modelled rather than averaged. The interface is white-labelled to JOTT's own codes, so the guidance feels like the brand lending a hand rather than a plug-in interrupting the sale. The result is a product page that resolves the layering question silently: the shopper answers once, and every jacket in the range knows their size from then on.

The numbers vindicated the approach. With Kleep live, JOTT saw conversion multiply by 2.8 among guided shoppers, hesitation at the size selector converted into decisions, while returns fell by 24%, easing the reverse-logistics load that weighs heaviest on bulky outerwear. For a brand whose margins live and die on seasonal sell-through, fewer boomerang parcels and more first-time-right deliveries compound quickly. JOTT's pattern matches what Kleep now measures across a portfolio of more than 300 live brands, where deployments typically cut size-related returns by 20 to 50% and lift assisted conversion by 15 to 25%, but outerwear remains one of the categories where the effect is most visible, because the cost of a wrong guess is highest. A jacket that packs into its own pouch now ships with something equally compact: the right size, first time.