Chantelle’s most consequential technology decision in recent years never reached a customer. The French lingerie group spent that time digitising its upstream: the suppliers, the purchase orders and the product data behind its nine brands. The bet underneath that work is that a manufacturer recovers its margin behind the factory door, in the half of the chain no one photographs, well before a shopper reaches the product page. Much of the industry has run the sequence the other way, aiming its AI budget at the customer-facing layer, from styling engines to virtual try-on to the zero-click content surfaces we covered this morning. For a group whose products are this complicated, Chantelle’s order of operations is harder to dismiss than it first looks.

The margin lives upstream

The complexity that decides Chantelle’s margin lives upstream, where almost no technology budget usually goes. A single bra can require more than 80 different materials, and multiplied across colours, cuts and sizes that math leaves the group carrying close to 80,000 active SKUs. That is not a marketing problem; it is a coordination problem with a marketing bill attached. Until recently the coordination ran on what the supply-chain vendor e-SCM Solutions calls “scattered tools”: spreadsheets, email threads and calls to suppliers. Every fragmented exchange is a place where a component lands late, in the wrong quantity or the wrong cup size, and in this business late stock becomes markdown stock. The leak is upstream; the loss shows up at the till.

What Chantelle built in response is a piece of plumbing, not a product. Through e-SCM, a collaborative platform for fashion supply chains whose rollout the group’s managers walked through publicly, it pulled order tracking, production status and supplier exchanges onto one system with end-to-end visibility. “What we were looking for was end-to-end visibility of our flows,” the group’s supply and logistics director, Jean-Michel Delperie, told the vendor; “we needed a collaborative, real-time tool to ensure data reliability.” The more revealing line comes from central planning manager Marta Pijewska, who says the automated reporting makes it “easier to anticipate cash outflows and manage inventory in transit.” That is an income-statement sentence wearing a logistics uniform. Visibility into inventory in transit is visibility into working capital, and working capital is where a manufacturer founded in 1876 either funds its next collection or finances dead stock.

The layer no shopper sees

Underneath the platform sits an even less photogenic layer. To make any of it work, Chantelle had Semarchy connect its SAP, Cegid and Cylande systems through more than 1,500 interfaces, the integration work that lets one inventory number mean the same thing in every brand and channel. It also added Centric’s product-lifecycle system to manage that SKU sprawl, a tool the group conceded it had run without for two decades while relying on an ageing product-data database. Centric has said the system will eventually be “rolled out to some of Chantelle’s suppliers,” which is the quiet tell in the whole programme. A lingerie group does not push its own product system into its factories unless it has decided the supplier, not the shopper, is the relationship worth instrumenting first. None of this is visible from a storefront; all of it changes what reaches the shelf, when it arrives, and what it costs.

A better product page has never once sold a bra that reached the warehouse two weeks late in the wrong cup size.

, each with a small supplier tag, centred and fanning outward against a plain ground, clinical and precise mood}})

The storefront has a ceiling

The rest of the industry has spent the AI cycle optimising the opposite end of the chain. The money and the headlines have gone to the customer-facing storefront: conversational styling, virtual try-on, recommendation engines, and the zero-click content surfaces that now swallow whole marketing budgets. Those tools are real and some of them work, but they share a ceiling. They raise conversion on the margin a product already carries; they do not change that margin. In a catalogue this size the recoverable money sits in residual stock, lead times and the working capital trapped between factories in North Africa and the 10,000 points of sale that finally put the product on a shelf, none of which a styling engine touches.

Where the thesis could break

The strongest objection to all this is that the numbers proving it are the vendor’s, not Chantelle’s. e-SCM reports a +20% productivity gain in the Chantelle case study and a -12% reduction in residual stock from its platform, but both are the supplier’s marketing figures, and Chantelle has published no audited margin line of its own to confirm them. The productivity figure comes from the Chantelle case study header; the residual-stock figure is drawn from e-SCM’s generic platform marketing and is not attributed to any single client. Grant the objection its full weight: if customer-facing AI turned out to lift margin more reliably than back-office integration, then upstream-first would be a laggard’s alibi rather than a strategy. But storefront gains are bought from the same handful of vendors every competitor uses, so they get competed away almost as fast as they appear. Upstream gains are entangled with one company’s particular suppliers and product mix, so they compound and resist copying. And Chantelle’s own sequence corroborates the direction, because a group that doubted the margin was upstream would not have spent years and three platforms there before building anything of comparable ambition for the customer.

That sequence is the lesson other groups can still choose to ignore. If the recoverable margin in apparel sits upstream, then a brand that pours its AI budget into the storefront while its suppliers still trade on spreadsheets is optimising the wrong half of its own business. The cost of that choice is not a worse website. It is a structurally lower margin that no conversion rate will repair, because the loss was booked long before the customer arrived. Chantelle rebuilt the half of the chain nobody photographs, and in apparel that, not a smarter storefront, is the half that reaches the accounts.