308 new dark stores in one quarter against 69 is the shortest version of this story. As of November 10, 2026, the public record shows Eternal's Blinkit adding 308 net stores in the September quarter while Swiggy's Instamart added 69, taking its network to 1,240 outlets across 133 cities Urban Acres. Both numbers describe capacity, and neither says whether the shelf behind each new door can hold the range it lists. That is the number a store network should settle before it signs the next lease.
The network grew faster than the margin
The quarter produced two numbers that usually move together and this time did not. Order value rose 34.8% to Rs 6,594 crore, while the adjusted core loss widened to Rs 895 crore from Rs 849 crore a year earlier Urban Acres. Revenue growth of that size without margin repair is the signature of a network still buying volume, and store count is the cheapest part of that purchase as well as the easiest to report.
Instamart's gross order value rose 31% year on year to Rs 9,230 crore in the same quarter, with monthly transacting users at 14.8 million, up 23% Urban Acres. More users and more orders with a thinner contribution is what a denser but not yet productive network looks like from the outside. The company said store utilisation and retention would support future margin improvement, which is a statement about availability rather than about marketing.
What 1,240 doors actually prove
A dark store is a promise with a lease attached. The promise is a delivery window measured in minutes; the lease is signed for years. Between the two sits the only thing that decides whether the promise is kept, which is whether the item a customer ordered was on the pick face at the hour they ordered it.
Density shortens the ride, not the pick
Adding nodes reduces the distance a rider travels, and it does not reduce the chance that a picker walks to an empty shelf. The same catalogue is being spread across more locations with less depth per location, so the arithmetic of availability gets worse before it gets better. Networks that expand quickly tend to cut range per node to protect the fill rate, which improves the reported number while quietly shrinking the assortment that made the channel attractive.
That trade-off is invisible in quarterly reporting, and it shows up later in two stages. First the basket keeps missing one item, and then the customer stops opening the app for the weekly shop and only uses it for emergencies. By the time the second stage appears in the numbers, the node has been open for a year and the lease has three more to run.
Availability is counted in the wrong place
Most retail organisations report availability from system stock, which counts cases in the back room and stock on a neighbouring shelf as sellable. A shopper can buy neither, and the gap between system stock and what an auditor finds on the pick face is the part a ten-minute promise cannot absorb. There is no time to substitute, no time to re-route and no second chance inside the same basket.
Why the metric has moved into investor decks
Availability used to sit inside supply-chain reviews, and it now appears in investor presentations because it is the only figure that connects a dark store to a shelf. When a network adds doors while pushing online share higher, availability has to be measured per node rather than per banner. A figure that averages a flagship store and a new suburban node tells you nothing about either of them.
Two markets, one constraint
The same constraint is visible in Brazil, where capacity was announced before the promotional calendar rather than after it. Shopee opened four new distribution centres in the country and tripled the operating area of one São Paulo facility, adding roughly 2.4 million parcels of daily processing capacity for 30 centres in total and more than 60% more order capacity than last November Ebrun International. The same announcement included plans to hire more than 14,000 people, mostly in logistics, because a building does not pick an order.
Mercado Livre is investing 57 billion reais in Brazil, taking large warehouse sites to 72 before Black Friday and offering 60-minute delivery in parts of São Paulo Ebrun International. Read together, the two moves describe a market where the discount calendar is now the easy part and the network is the constraint. A merchant that sets a discount depth without knowing the processing capacity behind it is selling a promise it does not control.
| Market | What was added | What it actually tests |
|---|---|---|
| India, dark stores | 69 net new nodes at Instamart; 308 at Blinkit in one quarter | Whether each node can hold its listed range all week |
| Brazil, fulfilment | Four new Shopee centres and a tripled São Paulo site | Whether dispatch keeps pace when the discount calendar runs a month |
| São Paulo, 60-minute zone | Sub-hour delivery inside a defined radius | Whether inventory is positioned where the promise is made |
The 60-minute zone is best read as a coverage claim rather than a national service level. Sub-hour delivery only works inside a radius where stock is already positioned, so the experience depends on where a shopper lives rather than on which platform they chose. For a brand, that turns speed into a distribution question, because the SKUs held inside the qualifying radius are the SKUs that can win the fast order.
A three-number gate before the next door
Three numbers are enough to start because they fail differently, and each failure points to a different fix. A node that audits at 95% on opening day and settles at 82% has not added capacity; it has added a promise it cannot keep. A node with a 91% fill rate and a 3.4% stock-cancellation share is not fixed by adding riders, and a node with both numbers healthy but a shrinking assortment is a merchandising decision wearing an operations costume.
FAQ
Does a negative contribution margin mean the channel is broken?
Not on its own. It means each order contributes less than the cost of fulfilling it before fixed costs. The question is whether mature nodes are positive and new nodes are dragging the average down.
Why measure availability per node rather than per city?
Because the promise is made per address. A city average hides the node that opens at 95% and settles at 82%, which is exactly the node that should not have opened.
Is a wider range always better for availability?
No. A wider range with the same depth per SKU raises the chance of an empty pick face. Range and depth have to be decided together, per node.
What does the Brazil expansion add to the picture?
It shows the same sequence in a different format: capacity announced before the discount calendar. Discounts get the headline; capacity decides who can honour them.
