Quick commerce revenue rose 45% to RMB53.3 billion in the June quarter while adjusted EBITA in the same segment fell 84%, and as of November 9, 2026 the public record still stops short of the number operators need most: the contribution margin per delivered order. Growth of that size, funded at that cost, turns a logistics question into a store-network question — how many locations can serve a three-kilometre ring before picking and rider time eat the basket.
What Alibaba's June quarter actually reported
Two numbers carry the story, and they point in opposite directions. Revenue of RMB53.3 billion in a single quarter is a demand signal: Chinese shoppers will pay for delivery inside the hour, often enough to build a line item that size. The 84% fall in adjusted EBITA is a supply signal: every order inside that number is served by a store, a picker and a rider, and none of the three gets cheaper because volume went up.
Read together, they describe a category that has already won the demand argument and has not yet won the cost argument. The category growth is no longer in question, and neither is consumer demand: the June quarter settled both. What remains open is which store format can serve the ring at a cost the average basket can carry. Alibaba's USD 1.5 billion offer for Pupu, disclosed while regulators were still reviewing the deal, suggests the answer is being bought rather than built store by store.
Where the store network absorbs the cost
Dark stores versus shop-floor picking
A dark store concentrates picking in a space nobody shops in. Shelves are laid out for a picker's route rather than a shopper's eye, and a single order can be assembled in a few minutes. The model pays for that speed twice: once in rent on space that generates no walk-in trade, and again in the inventory it has to hold to keep on-shelf availability high across a wide assortment.
Shop-floor picking spreads the same work across an existing store network, so the rent is already paid by the retail business. The constraint moves to availability. If the item is not on the shelf, the order ships short, and short orders are the ones that turn into refunds and one-star reviews. That is why on-shelf availability, not delivery speed, separates a working O2O network from an expensive one.
Rent, riders and the catchment radius
Every fulfilment model is a bet on density. A three-kilometre ring around a store holds a certain number of households, and those households place a certain number of orders each week; divide the fixed cost of the site by that order count and the cost per order falls out. Widen the ring and more households come in, but rider time per order rises faster than the basket does.
Rent is the fixed half of that bet and rider time the variable half, which is why two chains with identical revenue can report very different margins. A chain that adds stores faster than order density grows is buying revenue at a negative contribution. A chain that waits for density cedes the ring to a competitor who moved first.
Three fulfilment models, one margin question
| Model | Catchment | Fixed cost | Where margin leaks | Signal to watch |
|---|---|---|---|---|
| Dark store | 1–3 km, dedicated site | Rent on space with no walk-in trade | Picker idle time between order waves | Orders per square metre per day |
| Shop-floor picking | Existing trade area, up to 3 km | Shared with retail rent | Stockouts that cut the basket | Availability by top-selling SKU |
| Marketplace delivery | City-wide, pooled riders | Platform fee per order | Rider wait at pickup | Delivery time variance |
The three models are not ranked here, because the ranking depends on density and density is local. What the table does show is that each model leaks margin in a different place, and each leak is measurable. An operator tracking only revenue growth will read all three as equivalent. An operator tracking availability, idle time and delivery variance will see which one is working in a given trade area.
The June quarter sits on top of this. A 45% revenue increase against an 84% EBITA decline is what a category looks like when it adds capacity ahead of density: the capacity is real, the density is the bet, and the fiscal 2029 profitability target is management's own estimate of how long the bet takes to settle.
What an O2O team should log every month
The monthly scorecard below is deliberately short, and it is built to be read by the people who schedule pickers rather than by the people who write quarterly decks. A metric nobody can act on is worse than no metric at all, because it absorbs attention and changes nothing. Each of the four entries maps to a decision a store or category manager can take inside the month, which is the only test that matters for a network adding sites this quickly.
- Count stores by catchment type, not by city, so that a suburban ring and a downtown ring are never averaged together.
- Measure on-shelf availability on the assortment that actually drives the basket, then trace every short order back to its missing SKU.
- Record pick time and rider wait separately, because the two fail for different reasons and are fixed by different teams.
- Price the fixed cost per order against the order count of the same ring, and revisit the next site decision only after that ratio is stable.
| Field | Example value | Reading |
|---|---|---|
| On-shelf availability | 92% of the top 500 items in the ring | Below 90% the basket starts shrinking |
| Orders per store per day | 1,400 | Compare within one catchment type only |
| Pick time per order | 4 minutes 20 seconds | Rises as the assortment widens |
| Stockout-driven refund rate | 1.8% of orders | Usually the cheapest number to move first |
Logged together at store level, the four fields hold a picture that survives a promotion. Logged as city averages, the differences between trade areas vanish exactly in the month they matter, and the next site decision gets made on the number that flatters the network rather than the one that explains it.
FAQ
Does a 45% revenue increase mean quick commerce is profitable?
No. Revenue growth and segment profitability are reported separately, and management's target for the segment is fiscal 2029. Growth shows that consumer demand exists; margin shows whether the store network can serve it without subsidy.
Why does on-shelf availability matter more than delivery speed?
A missing item cuts the basket immediately and surfaces in reviews, while two extra minutes of delivery time is invisible to most shoppers. Availability is also cheaper to fix, because it is a stocking decision rather than a capacity decision.
Is the Pupu deal part of the same story?
It reads that way. Buying an existing dark-store network reaches density faster than opening sites one at a time, and the regulatory review is the timetable to watch before the deal can be treated as settled.
How many stores should one city have?
Enough that each ring reaches order density before the next site opens. The number differs by city; the test, orders per store per day measured against household count in the same ring, does not.
