As of November 7, 2026, Zalando's third-quarter release describes a change that goes beyond a normal marketplace expansion. The company reported group gross merchandise volume up 7.5 percent to 4.54 billion euros and revenue up 5.1 percent to 3.17 billion euros, then introduced Zalando Extended, a service that adds products from other online shops which Zalando does not sell itself. From 2027, Zalando's own AI agent will complete those purchases on the customer's behalf inside the Zalando app, with Westwing as the first pilot partner. The listing is no longer the last step a brand controls.
The checkout moves inside someone else's app
The structure matters more than the launch date. A partner shop does not integrate into Zalando's systems in the usual sense: the AI agent does the work, so the shop reaches Zalando's customers without complex technical integration or changes to its logistics. In exchange, the shop hands the customer relationship, the checkout flow and the data generated along the way to the platform. That is a different trade from a classic marketplace listing, where the merchant at least owns the transaction page.
Four numbers that move when a third party checks out inside the app
The commercial logic becomes clearer when the same quarter's numbers are read together. Gross margin fell to 38.7 percent from 39.6 percent, which Zalando attributed to consumer retail offers and faster growth in the lower-margin B2B segment. Net result swung to a loss of 18 million euros from a 12.5 million euro profit. At the same time, B2B revenue grew almost 28 percent to 354.4 million euros, partner business reached 33.6 percent of consumer GMV against a 2028 target of 40 to 50 percent, and retail media grew nearly 18 percent. An extended assortment fed by partners fits that mix: it adds selection without adding inventory.
| Metric | Q3 2026 | Why an agent channel moves it |
|---|---|---|
| Gross merchandise volume | 4.54 billion euros, up 7.5 percent | Third-party items add volume the platform never stocks |
| Gross margin | 38.7 percent, down from 39.6 percent | Partner sales carry a different margin structure than own retail |
| Partner share of consumer GMV | 33.6 percent | Extended assortment pulls more volume through partners |
| Retail media revenue | Up nearly 18 percent | Partner shops can buy visibility inside the same app |
What a partner shop gains and gives up
Reach without integration
The pitch to a partner shop is straightforward: access to 63.2 million active customers and a trusted app, without building a new channel or reworking fulfilment. The Zalando Assistant, which more than 22 million unique users visited this year, already sits in the discovery path, so the jump from being discussed in a conversation to being purchased is short. For a home and living brand with bulky items and slow turnover, that reach is hard to buy elsewhere at a comparable cost.
The price of the shortcut
What the shop gives up is control over presentation and over the customer record. If the agent assembles the basket, the shop cannot upsell at checkout, cannot run its own post-purchase flow, and may not see the search terms that produced the order. Retail media inside the platform becomes the only lever for visibility, which turns a channel decision into a media budget decision. That is a reasonable trade for some categories and a poor one for brands whose margin depends on attachment rates.
Three checks before joining an agent channel
For a brand or a mid-sized retailer weighing a similar arrangement, the useful preparation is not a technology project but a short list of commercial questions answered in writing before the first integration call. Consumer demand for agent-assisted buying is still small, yet the category growth in agent-referred traffic is real enough that a late start costs more than a careful one. Three checks cover most of the risk.
The second check is the one that gets deferred, and it is the one that decides whether the channel is profitable. Returns in home and living are expensive and often involve two people and a van; an agent that closes the sale without carrying that cost shifts a real expense onto the partner. Agreeing the split before volume arrives is easier than renegotiating it afterwards.
Where the model is still unproven
Three things remain open. The first is trust: an agent that buys on a customer's behalf has to be right about size, colour and compatibility, and home goods are exactly where consumers expect to be wrong once. The second is scale: one pilot partner and a 2027 start date is a test, not a channel. The third is economics, because the model only works if partner shops can absorb returns and still pay for visibility. Until those are answered, the most useful reading for a brand is that product data quality has become a distribution question, and the brands that publish clean price, stock and specification data will be the ones an agent can actually buy from.
FAQ
What exactly is Zalando Extended?
It is a service that adds products from other online shops which Zalando does not sell itself to the Zalando platform. From 2027, Zalando's own AI agent will complete the purchase inside the app, so customers do not leave it, and partner shops reach Zalando customers without technical integration or changes to their logistics.
Why does this matter for a brand that already sells on marketplaces?
Because the checkout, the customer record and the visibility lever move to the platform. A partner shop gains reach but gives up control over presentation and post-purchase contact, and retail media inside the platform becomes the main way to be seen.
Which numbers should be watched after launch?
Partner share of consumer GMV, gross margin, retail media revenue and return rates on extended categories. The first three are already reported; the fourth is the one that decides whether the model is profitable for partner shops.
Is the category ready for agent checkout?
Not fully. Size, colour and compatibility errors are common in home and living, and one pilot partner with a 2027 start date is a test rather than a channel. Clean structured product data is the prerequisite before any of it scales.
