Veepee no longer trusts the closeout to a human eye. The company that pioneered the European flash sale is rebuilding it on two things the original model did without: second-hand stock, and an algorithm to run it. The bet is that the closeout survives only if software, not a merchandiser, decides what resurfaces and to whom.

Once, that eye was the whole product. Vente-privée, as the firm was called in 2001, sold the timed reveal: a merchandiser staged brand overstock as an event, and scarcity did the selling. What resurfaced, and when, was a question of taste.

The flash sale is now the slower part of the business. Group turnover reached €3.3 billion in 2024, barely above the year before, on 82 million products across 22,000 sales events. Its Re-turn service runs member-to-member resale of used goods. The growth is in the wardrobe, not the closeout.

Granjon knew the model worked. “Vinted is a real success,” he said in 2021, “but we were not about to copy the model.” Veepee built one regardless.

His remedy for running it is software. Veepee frames artificial intelligence as a “holy trinity” of predictability, optimization and relevance, and sells resale as a win for the planet and the young shopper. That second claim is not a climate policy — it is an inventory strategy wearing a conscience.

The colder reason for the machine is the inventory itself. Resale is a warehouse of single copies, each item sold once and never restocked. It is the hardest catalogue a recommender can face, and the last a merchandiser could price by hand.

The flash sale was a merchandiser guessing what you had not thought to want; the recommender is built to have already decided.

Veepee is betting that relevance can do the work surprise once did. Its oldest customers came for the surprise.