Editorial Article
The Per-Store-Per-Day Benchmark Does Not Predict Profit, and India Proves It
India’s three quick commerce platforms disclosed their store counts, their throughput and their profits in the same quarter. The busiest network lost the most money. It is the cleanest test the per-store-per-day benchmark has ever had, and the benchmark fails it.
The quarter to 31 March 2026: Zepto ran 1,139 dark stores producing INR 784,764 of order value each per day. Blinkit ran 2,243 producing INR 704,806. Swiggy Instamart ran 1,143 producing INR 545,605. Adjusted EBITDA came in at minus INR 1,247.5 crore for Zepto, plus INR 37 crore for Blinkit, minus INR 858 crore for Instamart.
The operator with the highest revenue per square metre of dark store is the one with the largest loss. That is not a paradox and it is not a data problem. It is what happens when a single ratio is treated as a proxy for a two-variable business.
Where the benchmark comes from, and why it usually works
The logic behind per-store-per-day is sound as far as it goes. A dark store has a fixed cost — rent, refrigeration, base staffing, the software that runs the pick path — that does not move with order count. In a business with a fixed cost, revenue per unit of fixed capacity is the thing that decides whether you cover it. That is why the metric became standard across the sector and why Pheonix Research applies it to every market in this cluster.
What the Indian disclosures add is the missing variable. Revenue per store per day tells you how much fixed cost you are covering. It says nothing about how much gross profit is inside that revenue, and gross profit per order is set by the basket.
The second variable
| Q4 FY26 | Order value per store per day (INR) | Net order value (INR) | EBITDA per order (INR) |
|---|---|---|---|
| Zepto | 784,764 | 387 | −59.40 |
| Blinkit | 704,806 | 525 | +1.35 |
| Swiggy Instamart | 545,605 | 504 | −76.20 |
Zepto’s fixed cost per order was INR 62.67 in FY26, down from INR 80.19 the year before — a real operating improvement, and the throughput figure reflects it. But that rupee of cost is competing against a INR 387 sale at Zepto and a INR 525 sale at Blinkit. The same cost, absorbed by a third less revenue per order.
The pair is what decides: revenue per store per day sets the capacity you bought, and basket sets what each unit of that capacity earns. Either one alone will mislead you, and the sector has spent four years quoting the first.
Throughput can be bought
There is a harder point underneath this. Order count in quick commerce is not an independent measure of operating quality, because it responds directly to price. Discount the order and you will move more of them through the same site. Zepto’s annual transacting users went from 10.57 million in March 2024 to 38.38 million in March 2025 to 47.97 million in March 2026, funnelled through a network that did not grow at anything like that rate — which produces exactly the throughput number the benchmark rewards, and exactly the loss the P&L shows.
A metric that rises when you spend money to make it rise, and that is silent about the money you spent, is an incomplete metric. The per-store-per-day figure is a revenue ratio dressed as an efficiency ratio.
What survives of the benchmark
The benchmark is not useless; it is conditional. At a given basket, throughput determines whether a store covers its fixed cost, and that remains the single most useful operational question in the format. The correction is to never read it without the basket beside it, and to treat the USD 9,312 per node per day global average in this report as what it is — a revenue figure across 8,400 sites, not a measure of how well any of them is run.
| Benchmark | What it actually predicts |
|---|---|
| Order value per store per day | How much fixed cost is being covered. Nothing about margin. |
| Net order value (basket) | Gross profit per order, before fixed and delivery cost. |
| The pair together | Whether a site can clear its fixed cost — the actual question. |
Operator figures are for the quarter to 31 March 2026, from Eternal Limited’s and Swiggy Limited’s Q4 FY26 shareholders’ letters and Zepto Limited’s updated draft red herring prospectus dated 8 June 2026. Estate sizing and the per-format breakdown are in the Dark Store Networks & Micro-Fulfilment Market report.
Sources
The per-store-per-day order value and per-order profit figures compared here are computed from disclosed store counts, disclosed order value and the actual number of days in the quarter ended 31 March 2026:
- Eternal Limited, quarterly shareholders’ letters — Blinkit net order value, store count and orders.
- Swiggy Limited, financial results and shareholders’ letters — Instamart store count, order value and adjusted EBITDA.
- Zepto Limited, UDRHP-I filed with SEBI, June 2026 — store count, net order value and loss per order.
All three operators are taken from the same quarter deliberately: the estate is growing fast enough that figures from adjacent quarters are not comparable, and mixing them would change the ranking.
Frequently Asked Questions
Why does the busiest dark store network lose the most money?
Because the per-store-per-day figure is a revenue ratio, not an efficiency ratio. It tells you how much fixed cost a site is covering and says nothing about the gross profit inside that revenue, which is set by the basket. Zepto ran the busiest stores at INR 784,764 a day and lost INR 1,247.5 crore while Blinkit ran quieter stores at INR 704,806 and made INR 37 crore.
What are the three platforms' net order values?
INR 387 at Zepto, INR 525 at Blinkit and INR 504 at Swiggy Instamart in the quarter to 31 March 2026. Zepto's fixed cost per order fell to INR 62.67 in FY26 from INR 80.19 the year before, but the same rupee of cost is competing against a materially smaller rupee of sale.
Why is throughput not a measure of operating quality?
Because order count responds directly to price. Zepto's annual transacting users rose from 10.57 million in March 2024 to 47.97 million in March 2026 through a network that did not grow anything like as fast, which produces throughput the benchmark rewards and a loss the P&L records. A metric that rises when you pay it to rise, and is silent about the payment, is incomplete.
Is the per-store-per-day benchmark useless then?
Conditionally. At a given basket, throughput still determines whether a store covers its fixed cost, which remains the most useful operational question in the format. The correction is to never read throughput without the basket beside it, and to treat the USD 9,312 per node per day global average as a revenue figure across 8,400 sites rather than a measure of how well any of them is run.
