As of 13 November 2026, Swiggy's Instamart was running 1,240 dark stores across 133 cities after adding 69 net outlets in the quarter ended September, and the network was still being measured in doors rather than in profit. Gross order value rose 31 percent year on year to 9,230 crore rupees, net order value rose 34.8 percent to 6,594 crore, and monthly transacting users reached 14.8 million, up 23 percent. The adjusted core loss widened to 895 crore rupees from 849 crore a year earlier, and contribution margin slipped to negative 1 percent of gross order value from negative 0.2 percent in the prior quarter. Growth and margin moved in opposite directions, which is the normal shape of an O2O network before density catches up with footprint.
Doors added, margin given back
The disclosure gives a footprint, not a store-level picture. It reports 5.2 million square feet across 1,240 sites, which works out to an average size but says nothing about rent, lease terms, or how quickly an individual site reaches viability. For a brand or a supplier reading these numbers, the gap matters: a city count tells you where the network exists, while the question that decides whether your product sells is whether that node can pick it today.
The competitive frame is set by the same disclosure. Blinkit added 308 stores in the quarter to reach 2,751 and has said it wants more than 3,500 by March 2027, roughly four times the net additions Instamart reported. Two networks of different density are chasing the same urban demand, and the one that adds doors faster is also the one that must fill each new door with assortment, staff and stock before it earns anything.
Where the O2O maths actually breaks
Capacity triggers, not ambition
A new site is worth opening when the nearest existing site is full. That sounds obvious, but it changes the way a trade area should be scored: the question is not how many households sit inside the polygon, it is whether the site that already serves them is turning away orders. Scored the other way, expansion looks cheap right up to the point where two sites split the same demand and neither reaches break-even.
Cannibalisation nobody publishes
When a site opens within a kilometre of an existing one, the older site can lose a third of its daily volume while rent, staffing and equipment stay where they were. Platforms gain coverage across the city; the operator of the older site absorbs the drop. That asymmetry is why partner-run and platform-run stores should never be averaged together when judging network health, and why a store-level model needs three inputs before a lease is signed: saturation of the nearest site, overlap of delivery polygons, and the replenishment distance to the mother hub.
Replenishment is the input most often skipped. A site that runs out of its top sellers in the evening loses exactly the orders it was built for, because the shopper who needs milk at nine is not waiting until the morning van arrives. At 1,240 sites, a two-hour gap in the replenishment cycle at a few dozen nodes is enough to move the network's fill rate, and fill rate is the only number a brand can act on.
A four-row check before the next lease
The table below is the minimum diligence set for a new site. Values are illustrative; fill them with your own counts rather than city averages.
| Check | How to measure | What it decides |
|---|---|---|
| Saturation of nearest site | Orders per day against practical ceiling | Whether demand exists at all |
| Polygon overlap | Shared buildings between two sites | How much volume will shift |
| Replenishment distance | Km and hours to the mother hub | Evening stockout risk |
| Fill rate by node | Top sellers available at pick time | Whether listings convert |
The last row is the one that gets dropped when a network is expanding quickly. Fill rate by node is also the only row a brand can verify without access to platform data: order your own top sellers from three postcodes at three times of day and record whether the item was pickable. Coverage is a lagging indicator of where the network was built; availability is the leading one, and it is measured one node at a time.
FAQ
Does a bigger dark store network mean better availability?
Only if each node is stocked. Coverage without density forces assortment cuts, and the cut items are the first to break. That is why node-level fill rate, not city count, is the number to track.
Why is contribution margin negative while order value grows?
Because customer acquisition and new-site costs land before the density that pays for them. The margin improved year on year even as the absolute loss widened, which is the pattern of a network still buying coverage.
What should a supplier ask before listing?
Three things: how many nodes carry the item, what the replenishment cycle is at those nodes, and how the platform measures fill rate. Without those answers, a listing is a promise about coverage, not about availability.
Is the 2,000-orders-per-day trigger a rule or an observation?
It has been described as the point at which a new site is considered, which makes it a capacity signal rather than a guarantee. Treat it as a threshold to test against local data, not as a published target.
