Interactive field guide · Data through June 2026
The obvious story about quick commerce is speed. The more interesting question is what has to move for ten minutes to become routine: the goods, the planning, the journey and the work.
For most of history, households bought ahead of need, stored what they bought and remembered what was running low. The cupboard was a private warehouse, and every family ran one. Quick commerce proposes a different arrangement: let the network hold some of that stock nearby until it is needed.
01 · The system
By the time someone taps “place order,” most of the difficult decisions have already been made. The right goods must be sitting nearby. The store needs enough daily orders to pay its way. Pickers and riders have to be available at the same moment as demand.
Redseer estimates. Online grocery itself remains roughly 2% of total grocery retail. Source 1
Trace the order
A nearby dark store can carry safety stock that each household might otherwise hold separately.
Inference · Size unknownTrace enough orders and the same five things move every time: what’s in stock, who decided to buy it, how it travels, how much friction it took, and who ends up paying. The rest of this guide follows each of those in turn.
02 · Store economics
Blinkit, Instamart and Zepto are all large enough to make comparisons tempting. Their toplines, however, describe different things: Blinkit reports Net Order Value (NOV), Instamart Gross Order Value (GOV) and Zepto Net Revenue Value (NRV). None of the three totals settle it. What matters is the share of each company’s stores that already cover their own costs—the number Swiggy is the only one of the three to publish.
NOV rose 86.2% year on year. Adjusted EBITDA reached 0.6% of NOV. Blinkit’s move to inventory ownership changed what gets counted as revenue, so NOV gives the cleaner view of underlying growth.
Instamart ended the quarter with 1,171 active stores across 131 cities and 1.4 crore monthly transacting users. Monthly contribution breakeven arrived during the quarter; the full-quarter margin remained slightly negative.
Zepto ended FY26 with 1,139 dark stores. Adjusted EBITDA loss per order fell from ₹136.15 in FY25 to ₹78.75, even as the annual loss widened while the network expanded.
NOV, GOV and NRV are company-defined measures with different deductions and accounting models. Contribution margin also excludes central technology, employee, marketing and corporate costs. The cards preserve each label instead of manufacturing a league table.
On 11.5 crore orders, Instamart’s −0.2% contribution margin is roughly −₹1.38 per order ⊕. Its ₹778 crore adjusted EBITDA loss is about −₹68 per order ⊕. Zepto’s restated net loss works out to −₹92.24 across 640.18 million orders ⊕. The formulas appear with the sources below.
One order, several businesses
An order can generate product margin, customer fees and advertising revenue. The platform also controls assortment and can sell its own labels. The shopper buying detergent is one customer; the detergent brand bidding for attention is another.
Store threshold explorer
This is an explanatory model, not a reconstruction of any company’s P&L. Redseer’s mature-metro reference band is ₹500+ AOV and 1,200–1,400 daily orders per store. Product margin, incentives and cost allocation are not disclosed consistently.
03 · The household
Household shopping used to require anticipation. Someone noticed the detergent running low, held that thought until the next trip and bought enough to last. Quick commerce lets the decision wait until much closer to the moment of use.
Safety stock quietly ties up money, cupboard space and attention. A reliable nearby network can take on some of that burden. Frequent users clearly shop differently—smaller baskets, tighter timing between orders. Whether their cupboards actually hold less has never been measured directly.
Before cloud computing, companies bought servers sized for peak demand and kept spare capacity idle. Cloud providers pooled demand across many businesses. Ten thousand flats each holding emergency inventory is the same shape of problem, and a neighbourhood of algorithmically stocked dark stores is the same shape of answer.
The physical analogy has limits. Inventory cannot be provisioned instantly or duplicated at zero marginal cost. Perishables spoil, forecasting errors have a floor, and delivery capacity gets scarcer when demand spikes. Pooling works best where density is high, which is why this is a story about a small number of large Indian cities—not India as a whole.
Your pantry versus the network
Move the sliders and you can see how the trade-off would work. Confirming it for real would take a household inventory panel that does not exist yet.
Pre-paying for a delivery subscription can make customers order more often to “extract value” from the membership. Research on online grocery subscriptions finds that removing the per-checkout delivery fee fragments purchases: average order value falls, monthly spending can rise, and the retailer must pick, pack and ship more orders. The source and scope are documented below.
Spending atomisation
Both timelines total ₹30,000. One asks for four decisions; the other asks for forty. Smaller purchases may register less at the moment of spending. Whether that adds up to more spending overall is a question no published study has actually tested. Quick commerce may not be making Indian families financially reckless; it is making the monthly grocery budget a progressively worse description of how household consumption happens.
04 · New demand
Some orders would once have gone to a kirana or supermarket. Others begin as passing wants that would previously have been abandoned. We do not know how growth divides between the two, and that missing number changes almost every argument about the sector.
One late-night want
Whatever did not happen—the trip not taken, the purchase not made—was never logged anywhere. No public study has split the sector’s growth between displaced purchases and newly created ones.
Quick commerce tends to win
Kiranas still tend to win
India still has an estimated 13–14 million kiranas, and traditional trade retains roughly 85–90% of mass grocery. Branded and impulse orders may leave before low-ticket staples, taking a richer part of the basket with them. Nobody publishes kirana margins broken out this way. Until someone does, this stays a working theory, not a finding.
Quick-commerce adoption remains concentrated in “India 1”: roughly the top 10–15% of urban households. The model struggles further down the income distribution because packaged, branded staples carry a premium over loose commodities, while digital platforms cannot reproduce the informal credit—khata—and flexible quantities offered by local shopkeepers.
The result is selective democratization: convenience becomes widely available within the affluent urban segment, while the shopping and transit work is outsourced to a lower-paid gig workforce.
05 · The physical city
The promise only works when stock is already close. That means more neighbourhood stores, bulk deliveries arriving through the day, pickers working inside and riders leaving one order at a time.
Blinkit ended Q1 FY27 with 2,443 dark stores; Instamart listed 1,171 active stores at the same date; and Zepto reported 1,139 stores at 31 March 2026. Definitions and reporting dates vary. Blinkit is targeting 3,000 stores by March 2027 while Swiggy continues to expand Instamart’s network.
What did the rider replace?
The shopper’s journey had no motorised transport. The delivery adds a rider route and packaging, although routing and vehicle type still matter.
India-specific data on journey displacement, routing and incremental packaging does not exist at the granularity a sector-wide carbon number would need. So this guide does not offer one.
Labour
An app turns the customer’s trip into someone else’s paid route. The useful questions are earnings after vehicle and fuel costs, unpaid waiting time, income volatility, accident exposure, incentive rules, social insurance and a worker’s ability to challenge an algorithmic decision.
Blinkit’s adjusted EBITDA and Instamart’s contribution margin are calculated after operating costs, yet accident risk and income volatility remain substantially outside the platform P&L. Part of the ten-minute delivery’s price is settled somewhere other than the customer’s receipt.
06 · Where the work went
The app makes the purchase feel almost weightless. Behind it are shelves of stock, rent and electricity, inbound trucks, pickers, riders, traffic and packaging that has to go somewhere.
Some household safety stock may move into nearby stores.
Algorithms forecast demand as people replenish closer to use.
The shopping trip becomes someone else’s paid route.
Search, payment and fulfilment collapse into minutes.
Costs reappear in labour, retail margins, kerb space and waste.
What follows sorts that reality into three buckets: what the evidence actually shows, what it only suggests, and what nobody has measured yet.
India after instant
For a regular user, nearby stock can begin to feel like part of the home. When the network is reliable, a product held around the corner can seem almost as available as one in the next room.
The appeal is easy to understand. There is less travel and less need to plan ahead. A much wider range of goods is available at short notice. Delivering that convenience requires stock spread across neighbourhoods, constant replenishment, paid placement inside the app and a workforce organised around immediate demand.
For most of history, households answered the question “Who should hold the things we might need tomorrow?” with “we should.” Quick commerce offers another answer: “the network can.”
Nobody has run the study that would settle whether this makes the system more efficient, or just moves its costs somewhere harder to see. Until someone does, both readings stay on the table—and treating either one as settled is the actual mistake.
Company metrics are company-defined and are not necessarily comparable. Figures marked ⊕ are author calculations from reported data. Market figures are estimates, not official statistics. Last evidence review: 23 August 2026.