20,000 stores in 48 states, 608 supermarkets in Korea, 1,240 dark stores across 133 Indian cities. Three networks made the same move within weeks of each other, and none of them framed it as a bet on stores: the store stopped being only a place to shop and became a node in a delivery network. As of November 11, 2026, the numbers make the strategy legible. For category and channel teams the operational question has moved, from whether a store can ship an order to which store, in which trade area, is worth holding stock for.
Three networks, three answers
Korea is the cleanest comparison because both chains publish store counts for the same quarter. GS The Fresh added 27 locations while Lotte Super closed 31, widening the gap from 241 to 299. GS did it through franchising, which now covers 83% of its network, and the result reached the P&L: third-quarter revenue rose 14.9% to 527.7 billion won and operating profit jumped 159.8% to 23.9 billion won, both records for the company. Lotte took the other route, shutting low-margin outlets, and still lifted same-store sales 10% (Chosun Biz).
Franchise math, not store math
Franchising changes what a store is for. When headquarters controls assortment and ordering while the franchisee carries the rent, the marginal location is cheaper to open and easier to close, which is why GS could add stores in residential districts and new towns where the demand is close-to-home rather than destination. Quick commerce made that geography valuable: a denser network widens the area eligible for instant delivery, and GS reported quick-commerce sales up 52.2% in the quarter, with the channel reaching 10.1% of supermarket sales from 7.6% a year earlier.
What the Korean numbers say about payback
| Network | Stores at Q3 | Change in the year | Signal in the quarter |
|---|---|---|---|
| GS The Fresh | 608 | +27 | Quick commerce 10.1% of sales; revenue +14.9% |
| Lotte Super | 309 | -31 | Same-store sales +10%; operating profit +16.2% |
| Homeplus Express | 282 | 27 behind Lotte | New owner preparing regional openings |
| E-Mart Everyday | 248 | +7 from Q2 | Returned to net openings |
Read together, the four rows say something less obvious than expansion good, contraction bad. Lotte grew same-store sales 10% while its network shrank, and Homeplus is preparing to open in regional markets rather than defend the capital. Both moves assume the same thing: a store earns its place by serving a trade area it can actually reach, not by existing. The 10.1% share is the number to watch, because above that level pick-and-pack labour, stock accuracy and shelf availability stop being store problems and start appearing in the online P&L.
The 10.1% threshold
Below ten percent, orders fulfilled from store are an add-on that store labour absorbs without anyone noticing. Above it, the same payroll runs two businesses, and the failure mode is predictable: the shelf is empty exactly when the app says it is full. Korea's data does not yet show which chains solve that, but it does show the channel is no longer marginal to the store.
India shows the network itself is the product
Instamart's September quarter is the counter-case. Gross order value rose 31% to 9,230 crore rupees and net order value 34.8% to 6,594 crore, with sequential net order value growth accelerating to 13.4% from 2.5%. Monthly transacting users reached 14.8 million, up 23%, and the network added 69 net stores to reach 1,240 across 133 cities. Adjusted EBITDA loss widened to 895 crore rupees even as the margin improved to negative 9.7% of gross order value from negative 12.1%, and contribution margin narrowed to negative 1% (Urban Acres).
Density versus loss per order
A widening loss with an improving margin means the company is buying growth with fixed investment rather than with discounting alone, which is a different risk from the one dollar stores carry. The test is orders per store, not stores per city. A network that adds 69 stores and 23% more users is roughly holding orders per store flat, and that flat line decides whether the next 69 stores are funded by operations or by investors.
Dollar General bought reach, not exclusivity
The US case is the cheapest experiment of the three. Dollar General ran a pilot across roughly 7,000 stores in six states from late September, evaluated it for about seven weeks, and then went live nationwide across about 20,000 stores in 48 states with delivery in as fast as 30 minutes. It kept four delivery arrangements running at once rather than consolidating: Instacart, Uber Eats, its own myDG service and DoorDash. With 21,148 stores at the end of July, the chain is betting that reach beats exclusivity, and the timing adds a complication, since Instacart agreed this quarter to be acquired by DoorDash, which would put two of those channels under one owner (VTEX News).
What a store network test should ask
- Map the trade area before the store. Pull twelve months of delivered orders by postcode and check how many fall inside a plausible 20-minute radius of the candidate site, then compare that count with the catchment of the store you would cannibalise.
- Split the P&L in two. Run in-store revenue and fulfilled-from-store revenue against separate labour, packaging and stock-accuracy costs, because a blended margin hides which side is paying for the other.
- Set a shelf-accuracy floor. Measure on-shelf availability weekly on the top 200 SKUs and book a miss as a lost online order, not a shelving error, so the number lands in the commercial review instead of the store audit.
The third step is the one networks skip. Every case above depends on the same input, which is whether the shelf matches the system, and none of the three companies discloses it. That silence is where the next advantage sits: a retailer able to report shelf accuracy by store knows something its competitors can only estimate.
FAQ
Does a dark store network scale the same way as a supermarket network?
No. A dark store adds capacity inside a small radius and scales with order density per catchment, while a supermarket adds footfall and assortment. Mixing the two metrics produces a network that looks efficient on paper and fulfils slowly in practice, because store count and order volume are not interchangeable units.
Why did GS The Fresh grow through franchises rather than owned stores?
Franchising lowers capital per location and moves rent to the operator while headquarters keeps assortment and ordering. That structure makes the marginal store cheaper to open and easier to close, which matters when a network is being rebuilt around delivery coverage rather than destination shopping.
Is quick commerce cannibalising store sales?
Partly, and the split is measurable. Compare store sales inside and outside the delivery radius over the same weeks. If stores inside the radius fall faster than those outside, the channel is moving existing demand; if both grow, the network is adding orders it could not reach before.
What single metric should a category team watch on store networks?
On-shelf availability on the SKUs an app or assistant is most likely to surface. A store that is stocked on slow movers and empty on the top twenty fast movers fails the delivery promise while looking healthy on an inventory report.
