The Real Unit Economics of Quick Commerce After Platform Fees and Returns

Almost every brand we meet that is excited about quick commerce is excited about the wrong number. They quote GMV. They show a chart that goes up and to the right. They tell us Blinkit and Zepto and Instamart are now a meaningful slice of their topline. And then we ask the one question that changes the mood in the room. What is the contribution margin per order, after platform commission, after fulfilment, after the ad spend it took to get that order. Nine times out of ten, nobody has the number. The growth was real. The profit was a guess. And a guess that goes unchecked for two quarters is how a brand ends up scaling a channel that quietly loses money on every box it ships.

This is the core argument. Quick commerce GMV is not the same as quick commerce value. The platform sits between you and the customer and takes a cut at every layer, and unless you model the full per-order P&L, you are flying on a vanity metric. So let us actually build the number.

Start with the order, not the channel

The mistake is reasoning at the channel level. Total quick-commerce revenue minus total cost gives you a blended figure that hides everything. A few hero SKUs subsidise a long tail that bleeds, and the average looks fine while the structure rots underneath. The only honest unit of analysis is a single order, broken into its real components.

Take a representative order. Strip it down line by line. What the customer paid is the top line, and it is the last time that number flatters you. Everything after it is a deduction, and most operators have never written the deductions down in one place.

Platform commission. The take rate on the order value. This is the headline cut and it varies by category, but it is rarely small. It is the price of access to the dark store and the customer.
Fulfilment and handling. The per-order fee for storage, picking, and the last-mile that the platform runs on your behalf. It is roughly fixed per order, which means it punishes low average order values hardest.
Ad take-rate. The spend you put behind visibility, share-of-voice on the category page, and sponsored placement. Treat this as a cost of the order, because in a crowded dark store an unadvertised SKU often does not sell at all.
Returns, damages, and spoilage. Lower than ecommerce on many categories, but real, and brutal on anything perishable or fragile. A returned order is not zero revenue. It is negative, because you paid to ship it both ways and may not be able to resell it.
Your own landed cost of goods. The actual cost to make and deliver the unit into the platform’s network, including inbound freight.

What is left after all of that is contribution margin. Not gross margin. Not GMV. The rupees that actually remain to cover overhead and, eventually, profit. Most brands discover their contribution margin is far thinner than their gross margin suggested, and a meaningful share of orders are contribution-negative once ad take-rate is loaded in honestly.

The ad take-rate is the line everyone underweights

Commission and fulfilment are visible. They appear on a statement. Ad spend feels separate, a marketing decision rather than a cost of goods. That separation is exactly how brands fool themselves. On quick commerce the shelf is algorithmic, and discovery is bought. If you need to spend to win the slot, then the spend is not optional brand-building. It is the entry fee for that order to exist.

So load it into the per-order P&L. Take your channel ad spend over a period and divide it across the orders that period produced. Now your contribution margin tells the truth. We have watched brands realise that the SKUs they were proudest of were the ones being most aggressively subsidised by ads, and the apparent winner was a money pit wearing a growth costume. We go deeper on the availability side of this in our piece on why your Blinkit dark-store availability score matters more than your ad spend, because spending into a SKU that is out of stock at the dark store is the purest form of burning cash.

GMV measures how much money moved through the channel. Contribution margin measures how much money stayed with you. Only one of them pays salaries.

Where the order actually breaks even

Once the model is built, the break-even points stop being mysteries and start being levers. There are really only a handful of them, and every one is a decision you control.

Average order value

Fulfilment is roughly fixed per order, so AOV is the single most powerful lever in the model. A larger basket spreads that fixed cost across more revenue and can flip a contribution-negative order positive without changing anything else. This is why bundling, multi-pack architecture, and threshold nudges are not merchandising tricks. They are margin engineering.

Assortment

Not every SKU deserves a slot. The slow movers drag the blended number down, eat working capital, and often sit in the contribution-negative zone permanently. Disciplined pruning is one of the highest-return actions available, and we lay out the method in pruning slow movers as an assortment discipline. A tighter range that sells through is worth more than a wide range that mostly sits.

Take-rate negotiation and category mix

Commission is not always a fixed law of nature, especially as your volume grows. And category mix matters because take rates differ. A brand that understands its own per-order P&L walks into platform conversations with leverage, because it knows exactly which orders it can afford to chase and which it cannot. We get into how to actually run that conversation in our note on negotiating trade margin on quick commerce.

What changed recently

The take-rate side of this model is not standing still, and the move is in one direction. In March 2025 Blinkit shifted away from a fixed commission band of roughly 3 to 18 percent toward a dynamic structure where the rate is keyed to the selling price of items within a category, while Zepto’s take rate climbed to around 22 to 23 percent as it pushed to firm up unit economics ahead of a listing. Once storage, warehousing and delivery fees are stacked on top, the platforms’ combined share of the selling price now lands in the 30 to 35 percent range, with larger brands negotiating better terms, according to Business Standard. Instamart and Flipkart Minutes had not matched the commission changes at the time of that report, which is exactly why category mix and platform mix belong in the model.

The ad-take line is moving the same way. A Datum Intelligence projection cited by Storyboard18 puts combined Blinkit, Zepto and Instamart ad revenue at nearly Rs 4,900 crore for the current year, with an estimated 10 to 25 percent of FMCG and impulse-category performance budgets already shifting onto these platforms. More money chasing the same shelf means a higher entry fee per order, not a lower one. Inc42 reports that festive ad rates can jump 40 to 50 percent, and that the early return-on-ad-spend advantage normalises as keyword competition inside the apps rises. The practical reading for an operator is simple. If your model still assumes last year’s commission and last year’s cost-per-click, it is already optimistic. Rebuild it on current rates before you commit the next quarter of spend.

The honest comparison most brands avoid

Here is the question that sits underneath all of this. If contribution margin per order on quick commerce is thin after the full take-rate, is the channel even the right place for the next rupee of growth. Sometimes yes, because the velocity and visibility compound into brand value that a spreadsheet will not capture. Sometimes no, because a direct channel keeps far more of every sale and the brand is better served pushing there. That tradeoff is real and most brands get it backwards, which is why we wrote the marketplace versus D2C margin tradeoff as a companion to this.

The point is not that quick commerce is bad. It is that quick commerce is a channel with a specific economic shape, and you cannot manage what you refuse to measure. A brand that knows its true per-order P&L can scale the channel with confidence. A brand that knows only its GMV is gambling, and the house in this game takes a cut at every table, and is quietly raising the cut.

Build the model before you scale the spend

None of this requires a finance team or exotic tooling. It requires the discipline to write down every deduction against a single order and look at what survives. We build this model with every brand we run on quick commerce, because the alternative is scaling a number that feels like success and reads, in the accounts, like a slow leak.

This is the spine of Quick Commerce Growth as we practise it. Not chasing GMV for the deck, but managing contribution margin per order as the real scoreboard, and pairing it with the Marketplace Account Management discipline that keeps availability and assortment honest so the spend actually converts. Get the unit economics right first. Then scale the channel, knowing every additional order adds to the bottom line instead of quietly subtracting from it.

Which Cities to Launch Quick Commerce in First

Almost every brand we onboard arrives wanting to go live everywhere at once. The pitch deck has a map of India with pins in twenty cities, and the founder believes that a wide launch signals ambition. It does not. It signals that nobody has done the unit economics yet. Quick commerce does not reward national presence. It rewards per-store velocity in dense neighbourhoods, and that velocity lives in a surprisingly small number of metro clusters. Spread your launch across the country and you do not look big. You look thin, slow, and unproven, in front of the exact buyers who decide whether you get a reorder.

The discipline that matters most in the first ninety days is subtraction. Which cities you say no to, early, protects the cities that will actually carry your launch. Here is how we sequence that decision, and why metro depth has to come before any national dream.

Quick commerce economics are geographically concentrated

The thing to internalise is that quick commerce is not a national channel that happens to start in cities. It is a metro channel that may, later, extend outward. The model only works where order density, dark store density, and disposable income overlap. That overlap is real in a handful of clusters and thin to nonexistent everywhere else. The delivery promise that defines the channel depends on dark stores packed close together, and those stores only exist where demand already justifies them.

The numbers make the concentration explicit. As of late 2025, eight major metros (Delhi-NCR, Bengaluru, Mumbai, Chennai, Hyderabad, Pune, Ahmedabad and Kolkata) accounted for 68 percent of all dark store count in India, per a Savills India report carried by Storyboard18. Delhi-NCR alone carried roughly 400 stores, Bengaluru 360, Mumbai 250. That is the channel telling you, in its own infrastructure footprint, where it is deep and where it is not.

So the map you should be drawing is not a map of India. It is a map of roughly a dozen high-density metro pockets where the channel is genuinely deep. Everywhere else, the platform might technically operate, but your product will sit in stores that turn over slowly and read your SKU as a non-mover.

You are not launching in a country. You are launching in a handful of neighbourhoods. Win those completely before you let a single pin land anywhere else.

The tiered map that should drive your sequence

When we sit with a brand, we sort the geography into tiers before we talk about anything else. The tiers are not about prestige. They are about where a unit of your launch inventory has the best chance of moving fast.

The core cluster. The top metros where quick commerce is deepest, dark stores are densest, and the impulse-and-restock behaviour the channel runs on is already a habit. This is where almost every launch should begin, and where most should stay until the data is undeniable.
The secondary metros. Real demand, real depth, but a step behind on density. These earn their place only after the core cluster is performing, not alongside it.
The aspirational tier. Tier-two cities and emerging pockets where the platform operates but density is thin. These are a later-phase question, not a launch-day one. Putting inventory here early is putting capital to sleep.

The mistake is treating these tiers as a checklist to complete in parallel. They are a sequence. You graduate from one to the next on the strength of proof, and the proof comes from per-store velocity, not from how many pins are on the board.

Why national-first wastes spend and slanders your product

Launch inventory is the most expensive inventory you will ever place, because it is unproven and it sets your baseline. When you spread it nationally, you distribute it across hundreds of dark stores, and every store that holds your SKU has to sell it before you see velocity. In the dense core, that happens fast. In the thin tiers, it does not happen at all, and the dead stock there drags your blended numbers down.

The damage is not only financial. It is reputational, inside the platform. The category buyer reads your aggregate sell-through, and a national launch with thin coverage produces a weak aggregate even when your core cities are excelling. Your strong neighbourhoods get buried under your idle ones. You walk into the reorder conversation defending a number that a more disciplined launch would have made impressive. This is the same logic that governs picking your first quick commerce partner: concentrate your scarce inventory where your buyer is densest, so the early data flatters a good product instead of slandering it.

There is also a marketing cost. The promotional spend, the visibility slots, the ad budget that makes a quick commerce launch work all get diluted when they are stretched over twenty cities. The same rupees concentrated on three or four cities buy real share of shelf and real velocity. Diluted, they buy nothing memorable anywhere. And that ad budget is not optional any more. Retail-media spend on Blinkit, Zepto and Instamart is projected to reach roughly Rs 4,900 crore in 2026, per a Datum Intelligence estimate reported by Storyboard18. When the whole category is bidding for the same slots, spreading thin guarantees you lose every auction that matters.

How we actually prioritise the first cities

Demographic and category fit narrows the question, but the city call comes down to a short sequence of honest tests, answered with your real product and your real demand signal, not with hope.

Where does your existing demand already concentrate, from your D2C orders, your social audience, or your marketplace sales, so you launch into a warm pocket rather than a cold one?
In those pockets, how deep is the platform’s dark store density at the neighbourhood level, because city-level coverage hides enormous intra-city gaps?
Can your launch inventory generate strong per-store velocity in that cluster, or will it spread too thin even within a single city?
Is your product an impulse buy or a planned-basket buy, and does the cluster’s dominant shopper match that behaviour?
Can your promotional budget actually dominate share of shelf in these cities, rather than flicker across many?

Answer those well and the first-city list usually shrinks to three or four, sometimes fewer. That shrinkage is the point. The brands that win are rarely the ones that launched in the most cities. They are the ones that owned their first cities completely, then expanded from a position of proof.

City choice and store choice are the same decision

Picking the city is only half the work. Within each city, your inventory lands in specific dark stores, and which stores carry which SKUs is its own discipline. A great city with a sloppy store-level plan still produces weak velocity. This is where assortment planning by dark store stops being a nice-to-have and becomes the thing that decides whether your concentrated launch actually concentrates. It is also where pruning slow movers early protects the velocity number the buyer will judge you on. The city map and the store map are the same decision viewed at two zoom levels.

Sequencing the expansion after the core proves out

None of this means you stay in three cities forever. It means you earn each new city with evidence. Once the core cluster shows strong, repeatable per-store velocity and a healthy reorder rhythm, the secondary metros open up, and the case you make to the platform is backed by numbers instead of ambition. Expansion built on proof is cheap. Expansion built on hope is the thing that quietly drains a launch budget.

The other input that governs how fast you can expand is supply. You cannot widen your city footprint faster than your replenishment can keep the new stores stocked, and a stockout in a fresh city undoes the velocity you just paid to create. Getting this right is a forecasting problem as much as a geography one, which is why we run inventory forecasting for dark stores in step with the city sequence, never after it. The cities you can expand into are the cities you can keep in stock.

What changed recently

The platforms themselves are now pushing past the metros, and that shift changes the timing of your expansion decision, not its logic. As metro markets approach saturation, quick-commerce firms are pushing deeper into smaller cities for growth, with Amazon Now and Flipkart Minutes expanding into tier-1 and tier-2 markets, as reported by Business Standard. By the Savills data above, tier-2 and tier-3 cities already held around 32 percent of all dark stores by late 2025. It is tempting to read this as permission to go wide on day one. It is not. The platforms can afford to seed thin tiers because they are amortising fixed dark store cost across every brand on the shelf. You are amortising launch inventory across only your own SKUs, and a thin tier-two store still reads your product as a non-mover regardless of how aggressively the platform is expanding there.

The second change tightens the case for depth further. Platforms have steadily raised platform fees, handling charges and take-rates to fix their own unit economics, with brand-side commissions and fees now commonly cited in the 15 to 18 percent range before ad spend, per Storyboard18. Higher take-rates mean every dead unit in a thin city costs you more than it did a year ago. The expensive-inventory argument for concentration has only gotten stronger. When the channel takes a bigger cut, you cannot afford to feed it stock that will not move.

So the recent news cuts both ways, and the discipline holds. Yes, the map is widening. No, that is not your cue to widen with it on launch day. Let the platforms prove the thin tiers on their own capital. You concentrate on the deep core, bank the velocity, and follow the expansion only where your own numbers say the density now justifies it.

The launch that looks small and wins big

A disciplined quick commerce launch looks underwhelming on a map and overwhelming in the numbers. Few pins, deep velocity, a buyer who sees a product that moves. That is the launch that earns the next ten cities. The national-map launch looks impressive in the deck and arrives at the reorder meeting with thin coverage and a story it has to apologise for.

We run this city-prioritisation exercise before any listing goes live, because Quick Commerce Onboarding only pays off when the inventory is concentrated where the channel is deep. Done right, the same work feeds a broader quick-commerce-first FMCG launch, where geography, assortment, and supply move as one plan. Our Quick Commerce Onboarding and Marketplace Account Management teams treat the first city list as the most consequential decision of the launch, because it is. Choose depth over breadth. Win your few cities completely. Everything after that is easier and cheaper.

The Quick Commerce Margin Reality Check Before You Sign

A quick commerce onboarding deck is a beautiful thing. Reach across thousands of dark stores. Ten-minute delivery. A buyer who converts on impulse before the second-guess kicks in. The brand team comes back from that meeting energised, and somewhere a launch date gets set. What almost never happens in that room is anyone opening a spreadsheet and asking the only question that matters. After everything the platform takes, what is left on each unit. The honest answer is often nothing, and sometimes less than nothing.

This is not a reason to avoid quick commerce. It is a reason to model it before you sign, because the costs stack in a way that no single line item reveals. Trade margin looks survivable on its own. Fulfillment fees look survivable on their own. The ad commitment looks survivable on its own. It is the sum, applied to your actual unit, that decides whether you are building a channel or subsidising one.

The margin you agree to is not the margin you keep

The number that anchors every quick commerce negotiation is the trade margin. The platform buys from you at a discount to MRP and that discount is the headline cost everyone fixates on. It is also the easiest number to feel good about, because it is a single clean percentage and it is the one thing your team thinks it controls.

It controls less than it thinks. The trade margin is the entry fee, not the full bill. On top of it sit fulfillment and handling charges, platform or marketing fees that are often non-negotiable, payment and logistics deductions, and a return or damage allowance that nobody models until the first reconciliation. Each is small. Together they routinely add another large slice on top of the trade margin you shook hands on. We have written separately about how to negotiate the trade margin itself, but the trade margin is only the first of several conversations, and treating it as the whole deal is the most common mistake we see.

The trade margin is the price of admission. The fees are the price of staying. Founders sign the first and discover the second.

The fee stack, line by line

If you want to model this properly, stop thinking in one percentage and start thinking in a stack. Each layer is a deduction against the MRP your customer pays, and every layer compounds the one above it.

Trade margin. The platform’s buying discount. The headline, and the smallest part of the real cost in many categories.
Fulfillment and handling. Per-order or per-unit charges for picking, packing, and the actual ten-minute run. These scale with order volume, not with your margin, which is what makes them dangerous on low-ticket SKUs.
Platform and marketing fees. Often a fixed percentage framed as a cost of being listed. Frequently presented as non-negotiable, which means it has to be absorbed, not argued away.
Ad and visibility commitments. The spend you agree to so your product is actually findable inside the app. More on this below, because it is where unit economics most often die.
Returns, damages, and shrinkage. A real allowance, not a rounding error, especially in perishable, fragile, or impulse categories.
Payment and settlement deductions. Gateway costs and the working-capital cost of waiting weeks to get paid on goods you have already shipped.

Run those against a true cost of goods that includes inbound logistics and the packaging quick commerce demands, and the picture changes fast. A SKU that shows a comfortable margin in your D2C store can land at break-even or below once the full stack is applied. That is not a pricing failure. It is a modelling failure, and it is entirely avoidable.

The ad commitment is where unit economics quietly die

Here is the part the onboarding deck soft-pedals. Visibility inside a quick commerce app is not free, and it is not optional. The shelf is small, the buyer decides in seconds, and the categories above and beside yours are bidding for the same slot. If you are not paying for placement, you are functionally invisible, and an invisible SKU sells nothing regardless of how good your trade margin looks on paper.

So the ad spend is not a growth lever you switch on later. It is a cost of distribution you must price in from day one. The mistake is to model your economics at zero ad spend, agree to the deal, and then discover that the only way to move volume is to layer a meaningful ad rate on top of an already-thin margin. At that point the channel is not contributing. It is consuming. The decision to spend was made for you the moment you signed, and you priced it at zero.

This is one of the structural reasons quick commerce does not behave like a marketplace. On a large marketplace, organic discovery and search rank can carry a well-listed product for a long time. Inside a ten-minute app there is far less organic real estate to win, the assortment per store is deliberately narrow, and paid visibility is closer to mandatory. Importing your marketplace assumptions about free traffic is how the ad line item ambushes you three months in. If you are still deciding where to launch at all, our view on which platform to start with works through the same trade-offs platform by platform.

Model it per SKU, per store, before you sign

The blended view is the enemy here. An average margin across your catalogue will tell you the channel is fine while two hero SKUs subsidise a long tail that loses money on every unit. Quick commerce punishes this harder than most channels, because the platform decides which of your SKUs each dark store even carries, and it will not necessarily pick your profitable ones.

So the discipline is the same one we apply everywhere, taken down to the unit. Work out profitability one SKU at a time, with the full fee stack and a realistic ad rate loaded in, and you will usually find the channel is viable for a specific subset of your range and ruinous for the rest. That is a useful answer. It tells you what to actually list.

What the model needs to include

A defensible pre-signing model is not complicated, but it has to be complete. At minimum it should hold:

True landed cost of goods, including inbound freight and quick-commerce-grade packaging.
The full deduction stack above, not just the trade margin.
A realistic ad rate as a fixed cost of distribution, never zero.
A returns and damage allowance appropriate to the category.
The working-capital cost of the settlement cycle.
A per-SKU contribution line, so the losers cannot hide behind the winners.

If the contribution per unit is positive after all of that, you have a channel. If it is negative, you have a decision to make before you sign, not a surprise to absorb after. The difference between those two situations is one afternoon with a spreadsheet.

Assortment is the lever most founders forget they hold

The model will often tell you the answer is not yes or no, but which ones and where. A premium, higher-ticket SKU absorbs the fee stack far more comfortably than a low-margin impulse item, because the fixed per-unit fees become a smaller share of a larger price. The same logic applies geographically. Demand and margin both vary by dark store, and listing your full range everywhere is how the unprofitable combinations creep in.

This is why assortment planning by dark store is not an operational afterthought but a margin decision. The right move is frequently to lead with the SKUs that survive the stack, in the locations where they sell, and to keep the thin-margin tail off the channel entirely until volume or pricing changes the math. You hold this lever. The platform would prefer you list everything. Your model should decide, not their deck.

What changed recently

The fee stack has only hardened since this became standard advice, and the numbers are now public enough that no founder can claim surprise. Reporting in 2025 put effective platform costs at roughly 30 to 35 percent of revenue once listing fees, mandatory ad spend, commission, and operational charges are added together, with the working rule of thumb that the channel only pays for brands carrying gross margins north of 65 percent. That is the same arithmetic this piece has always argued, now confirmed at the line-item level.

The specific commitments are worth knowing before you walk into the room. Per Storyboard18, Blinkit has charged a mandatory listing fee of Rs 25,000 per SKU per state, credited to a non-refundable ad wallet that expires within twelve months, with monthly marketing spend on top running Rs 2 to 3 lakh. Instamart was quoted listing-cum-ad fees of Rs 8 to 10 lakh a quarter alongside fixed weekly product orders, and Zepto bundled ad slots, onboarding, and influencer marketing from Rs 5 to 6 lakh. In the same report, one seller described spending over a million in capital across these platforms in three months without clocking even 10 percent of expected sales, and return on ad spend for small brands was said to rarely clear 1.2 to 1.5 times. None of that shows up in a trade-margin negotiation. All of it lands in your contribution line.

The reason platforms lean on these fees is no secret either. Retail media is now the profit engine. A Datum Intelligence forecast cited by Storyboard18 projects Blinkit, Zepto, and Instamart will generate close to Rs 4,900 crore in advertising revenue in 2026, with industry estimates that 10 to 25 percent of FMCG digital performance budgets have already shifted to quick commerce. That demand is real, which is precisely why the ad commitment is not optional and why modelling it at zero is the costliest assumption in the deck.

What to do before the pen touches paper

None of this is an argument against quick commerce. The channel is real, the buyer is real, and for the right products it is genuinely additive. The argument is narrow and it is this. The margin you agree to in the room is not the margin you keep, and the gap between them is large, predictable, and knowable in advance. Model the full stack, load a real ad rate, run it per SKU and per store, and let the number tell you what to sign.

This is the unglamorous core of D2C & Marketplace Strategy Consulting, and it is the work that should happen before any onboarding call, not after the first reconciliation statement lands. Building the per-SKU contribution model, pressure-testing the ad commitment, and shaping the assortment so the channel pays its way is exactly where our Quick Commerce Management and Profitability & Unit Economics teams start. The platforms are not hiding the costs. They are simply not adding them up for you. That part is your job, and doing it one afternoon early is the cheapest decision you will make all year.