A dark store holds a few thousand SKUs, not a few million, so every pack you list has to earn a physical slot against everything else that could sit there. That favours small, fast, single-decision pack sizes over the large value packs that carry your basket on Amazon. Pick one hero SKU in the size a shopper buys without pausing, prove the margin works at that smaller ticket, hold availability in a limited set of stores, and only then earn the second SKU.
- →A dark store is a shelf with a fixed number of slots. Listing your full range does not make you available, it makes you slow in several places at once.
- →The winning pack is the one a shopper buys without stopping to compare. That is almost never the value pack that performs on a marketplace.
- →Margin has to be rebuilt at the smaller size, not assumed from the larger one. Fees come off a smaller ticket and packaging costs more per unit.
- →Rate of sale per store is the number that buys you a second slot. Total units across a city can hide a SKU that is barely moving in most of them.
- →Assortment is a cluster-level decision. The right range in a dense metro pocket is the wrong one two neighbourhoods away.
A Dark Store Is a Shelf, Not a Warehouse
Most brands arrive at quick commerce with a catalogue built for a marketplace and try to port it across. That instinct is reasonable everywhere else. On Amazon or Flipkart, listing more is close to free: a slow SKU sits in a fulfilment centre costing you a little storage, and the long tail eventually pays for itself. A dark store does not work that way. It is a small physical room serving a few kilometres, holding a few thousand SKUs against a marketplace catalogue that runs into the millions, and every slot in it is contested.
That single fact rewrites the assortment question. You are not asking what you would like to sell. You are asking what you can justify occupying space with, in a room where the category team can see exactly how fast the thing in that space is turning. A SKU that sells respectably on a marketplace and slowly here is not neutral. It is holding a slot that something faster wanted, and the platform notices long before you do.
The practical consequence is that a wide listing is not the same as being available. Brands routinely go live with eight SKUs, spread their stock across all of them, and end up partially out of stock on the one that was actually selling. On a ten-minute app an out-of-stock SKU is not a delayed sale, it is a lost one, because the shopper takes whatever is in the app right now and often does not come back to check. The mechanics of that, and what actually drives whether you get seen at all, are covered in more detail in how visibility really works inside a dark store.
So the discipline is subtraction rather than addition. Fewer SKUs, deeper stock on each, and a clear view of which one is the horse. That feels wrong to a team used to marketplace thinking, where breadth is a defensive asset. Here breadth is a tax you pay in slots.
Your Marketplace Hero Is Usually the Wrong SKU Here
The pack that carries your Amazon business is usually the large one, or the multi-unit combo, because a marketplace shopper is planning a refill and comparing on value per unit. They have already decided to spend twenty minutes on the decision. The quick commerce shopper is doing something else entirely: they have run out of something, they want it now, and the whole session lasts under a minute. Nothing about that state of mind favours a value pack.
What it favours is a pack the buyer can say yes to without thinking. Every category has a rough ceiling above which a shopper stops behaving impulsively and starts behaving comparatively, and once they cross it they open a marketplace tab or wait for the sale. You will not find that ceiling in a report. You find it by looking at what already moves fast in your category on these apps and noticing where the prices cluster, because the platforms have effectively already run the experiment for you.
The other half of the same point is frequency. Quick commerce is a replacement channel, so the pack size that suits it is the one people finish fastest, not the one that lasts longest. A size someone runs out of every three weeks generates a reorder every three weeks. A three-month pack generates one order and then goes quiet, which reads to the category team as a SKU that does not turn.
| Amazon and your own site | Blinkit, Zepto and Instamart |
|---|---|
| Large and value packs carry the basket value | Small and single-use packs clear the impulse ceiling |
| Full shade, flavour and variant range is an asset | Two or three SKUs, chosen for rate of sale |
| Bundles and combos raise average order value | Bundles slow the pick and rarely survive a dark store |
| Long tail sits in inventory cheaply until it sells | Every slow SKU costs a slot a faster one wanted |
| Shopper is planning a refill and will compare | Shopper is replacing something that just ran out |
This is also where a dedicated pack size earns its keep. Making a smaller SKU specifically for quick commerce costs an artwork run and a new code, and in exchange it stops your marketplace hero from being discounted down to meet an impulse price it was never built for. A separate pack is the cleanest way to keep two channels from arguing with each other, which is the same logic behind building a price architecture that survives selling in several places at once.
Rebuild the Margin at the Smaller Size Before You Commit
The most common way brands get hurt in this channel is by assuming the economics of the big pack carry over to the small one. They do not, and the gap is usually worse than expected, because several things move against you at once. Platform margin and fees come off a smaller ticket. Packaging cost per unit of product goes up, sometimes sharply, because you are paying for a similar bottle or pouch to hold less. Any per-order handling is amortised over less value. And promotional support, which the platform will ask for, is a percentage of a number that just got smaller.
None of that makes the smaller pack a bad idea. It just means the unit economics have to be worked out at the size you intend to sell, before you scale, rather than discovered in month four when the volume is real and the contribution is not. Build the sheet for the actual pack, with the actual deductions, and be honest about the promotional line rather than treating it as an exception you will negotiate away.
Then judge the result against what the channel is really buying you. If your category has genuine repeat behaviour, a thinner margin on a small pack can be worth it, because the second and third orders arrive without acquisition cost and the customer forms a habit around your brand being the one that is there when they run out. If your product is bought once or twice a year, that logic does not apply and the same money almost certainly works harder elsewhere.
"A slow SKU in a dark store is not a slow SKU. It is a slot you took from a faster one, and the platform is keeping score."
- Brand Integer Quick Commerce Team
Start With One SKU and Earn the Second
The sequence that works is narrow and slow at the start, which is uncomfortable when a category manager is asking for your range and the internal pressure is to look serious. Go live with one SKU in a limited set of stores. Keep it in stock without exception. Watch how it sells per store rather than in total. Only when that one SKU is holding availability and moving without constant discount support have you earned the right to add a second, and the second should be a variant of the thing that worked rather than a hedge against it.
Adding variants early is the trap, because a variant does not create demand, it splits it. Two shades or two flavours in the same store usually sell what one would have sold, across two slots, at half the velocity each, which is precisely the number the platform is judging. The exception is a genuinely different occasion or need state, where the second SKU is being bought by someone the first was not reaching.
Sequencing this way also gives you something to negotiate with. A brand that walks in with one SKU and a clean rate of sale per store has a fact to argue from when it asks for more stores or better placement. A brand that went live with eight and averages poorly across all of them has nothing to point at, and the wider conversation about how to actually build a channel plan across Blinkit and Instamart gets much harder to have from that position.
- Pick one hero SKU The pack a shopper buys without thinking, in the size they finish fastest.
- Check it against the impulse ceiling If the landed price makes a buyer pause to compare, the pack is too big.
- Rebuild the margin at that size Platform margin, listing fees and returns come off a smaller ticket, so the unit economics have to work before scale, not after.
- Prove per-store velocity Go live in a limited set of stores and watch rate of sale per store, not total units.
- Earn the second SKU Only add a variant once the first one holds availability and moves without support.
Assortment Is a Cluster Decision, Not a National One
The last mistake is treating assortment as something you set once for the country. Dark stores serve a radius of a few kilometres, which means each one is effectively selling to a single neighbourhood with its own income profile, household size and cooking or grooming habits. Two stores in the same city can want genuinely different packs, and the aggregate city number will average them into a picture that describes neither.
This is why total city volume is such a misleading signal for expansion. A city can look strong while carrying its numbers on a handful of dense clusters, with your SKU barely moving in most of the others. Expanding the whole city on that reading spreads your stock thinner in exactly the pockets where it was working, and the per-store rate of sale you were proud of quietly falls.
Work outward from clusters that resemble the ones already performing, and treat each new city as a fresh test with one SKU rather than a rollout of the range you assembled somewhere else. Larger packs, where they work at all, tend to work in a narrower set of places than the small ones do, so the range should widen only in the pockets that have earned it. The brands that hold this channel over a couple of years are not the ones with the biggest listing. They are the ones who know which pack belongs in which pocket, and are willing to keep everything else out.
Frequently Asked Questions
How many SKUs should a new brand list on Blinkit or Zepto?
Start with one, and plan for two or three at maturity in a given city. A dark store carries a few thousand SKUs against a marketplace catalogue that runs to millions, so the question is not what you would like to sell, it is what you can justify occupying space with. One SKU that holds availability and moves without discount support does more for you than five that each sell slowly, because the platform's category team judges you on rate of sale per store. Add the second variant only when the first is the constraint rather than the experiment.
Should we create a special pack size just for quick commerce?
Often yes, and it is usually the single highest-leverage decision in the channel. The pack that works here is the one a shopper buys without pausing to compare, which in most categories means a smaller size and a lower absolute ticket than your marketplace hero. Creating that pack costs a new artwork run and a new code, and it protects your larger pack from being discounted down to meet an impulse price it was never built for. Check the economics at the smaller size before committing, because platform margin and fees come off a smaller ticket and the per-unit packaging cost goes up.
Is quick commerce worth it if the margin is thinner than Amazon?
It depends on whether you are buying a sale or buying a habit. Quick commerce is a replacement channel: people order it when something has run out, which makes it strong for repeat purchase and weak for discovery and for anything a buyer wants to research first. If your category has genuine repeat behaviour, a thinner margin on a smaller pack can still be worth it because the second and third orders cost you nothing to acquire. If your product is bought once or twice a year, the same maths does not hold and the money is better spent elsewhere.
How do we decide which cities to expand our assortment into?
Follow rate of sale per store rather than total city volume. A city that looks large in aggregate can be carrying its numbers on a handful of dense clusters while most stores barely move your SKU, and expanding the whole city on that signal spreads your inventory thinner in exactly the places it was working. Look at how a SKU performs per store in the clusters where it is already live, expand within the clusters that resemble them, and treat a new city as a fresh test with one SKU rather than a rollout of the range you built somewhere else.
