पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 18 · Build
Selling on Amazon, Flipkart and quick commerce without losing margin
Marketplaces bring buyers who are already searching and take a share of every order. Decide which earn a place, compute the real fees per order and run their ads to a calculated break-even.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
A Jaipur home-decor brand lists a ₹799 cushion cover on two marketplaces and sells three hundred a month. Revenue looks healthy. At the end of the quarter the founder adds up the settlement reports and finds the brand made ₹11 on each cover after fees, returns and ads. The marketplaces made more on every order than the brand did. That was not the marketplaces’ fault. Nobody had done the arithmetic before listing.
For most Indian consumer brands the large marketplaces are where the first meaningful volume comes from, because that is where buyers already search. They are not a free channel and they are not a cheap one once every line is counted. This lesson is how to decide where to sell, how to compute what an order really earns, how to run advertising inside a marketplace without giving the margin away, and how quick commerce differs from both. The [D2C unit economics](/library/d2c-unit-economics-order-that-must-make-money) lesson does the same arithmetic for your own website.
Which marketplaces, for which job
Each marketplace rents you a different kind of buyer. Amazon and Flipkart bring national search demand: people who type a product or a problem and compare. They suit products people look for by name or category, with a price and a photograph that can win a comparison on a phone screen. Quick-commerce apps bring impulse and top-up demand in the large cities: people who want something in minutes and choose from a narrow shelf. They suit fast-moving products with repeat purchase and a brand a buyer recognises at a glance. Your own site brings the customers you have already won, at a better margin and with the data a marketplace keeps for itself.
Start with one marketplace, not all of them. Choose the one where your category’s buyers search, measured by looking at the bestseller lists and the number of reviews on the leading products in your category, and stay there until each listing makes money and ranks on the first page for the searches that matter. A second marketplace doubles the catalogue work, the stock planning and the reconciliation, and it should be added when the first is profitable, not when it is struggling.
The real fee stack
Every marketplace rate card has the same parts under different names. A referral fee or commission, a percentage of the price. A fixed or closing fee per unit. Shipping or weight handling, by weight band and distance. Fulfilment fees if the marketplace stores and packs your stock, plus storage. Kalaari Capital’s ONDC primer puts traditional marketplace commissions at 15 to 30 per cent depending on how much marketing and fulfilment is included. The marketplace also collects tax at source under GST and deducts income tax at source from your settlements; both are credits you can claim, but only if you reconcile them every month.
The parts move. On 1 March 2026 Amazon India announced zero referral fees on products priced under ₹1,000 across more than 1,800 categories, which it said covered over 12.5 crore products, from 16 March, extending an earlier phase covering 1.2 crore products under ₹300. It cut Easy Ship fees by more than 20 per cent for products under ₹300. Its own examples show what is left: a ₹999 necklace on Easy Ship saves ₹224 a unit, 69 per cent, and ₹798 earphones fulfilled by Amazon save ₹139, 56 per cent. By the announcement’s own percentages each still carries roughly ₹100 to ₹110 in fees per unit. Zero referral fee is not zero fees; on a low-priced product the fixed fees per unit are now the main cost, and they do not shrink with the price.
One decision changes the fixed fees more than any other: who stores and ships. Letting the marketplace hold your stock and fulfil orders usually costs more per unit in fees and storage, and often buys faster delivery promises and better placement for the listing; shipping yourself through the marketplace’s pickup service costs less per unit and leaves you holding the stock and the packing. There is no general answer. Run both through the figure below for your best-selling product, with the actual rate card for its weight band, and add what the faster delivery does to conversion once you have a month of data. Keep slow sellers out of the marketplace’s warehouse, where storage fees accrue on stock that does not move.
Returns, the cost nobody models
A returned order earns nothing and pays for delivery twice, once each way, and sometimes comes back unsaleable. Founders model returns as a small percentage of revenue when they are a cost per order shipped, and in fashion, footwear and home products they are often the line that turns a profitable listing into a loss. Measure the return rate per product from your own settlement reports, separate customer returns from cash-on-delivery orders refused at the door, and price the reverse leg at what the rate card actually charges. Then fix the causes you control: size charts, honest photographs, a description that answers the question the returns reasons keep giving, and packaging that does not arrive broken.
Advertising inside the marketplace
On a large marketplace the first page of results for most searches is partly paid, so advertising is not optional for a new listing; it is the price of being seen while the listing earns reviews and organic rank. Amazon’s own guide to advertising cost of sales defines the measure: ACOS is ad spend divided by the sales those ads produced, so ₹50 of spend producing ₹100 of sales is an ACOS of 50 per cent, and ROAS is the inverse. Its rule for profit is the one that matters: ACOS must be lower than your profit margin, and the place to start is the break-even ACOS, the margin before advertising.
Watch two numbers, not one. ACOS on a campaign tells you whether those ads paid. Ad spend as a share of all the product’s sales, including the organic sales the ads helped rank, tells you whether the listing as a whole is profitable, and it should fall month by month as reviews and rank build. A listing whose total ad share is not falling after three months is renting its sales, not earning them. The figure computes the break-even for one order.
A marketplace is a landlord with a crowd outside. Pay the rent only on listings where the crowd buys at a margin you have already computed.
Quick commerce is a different contract
Quick-commerce apps do not work like the large marketplaces. They hold stock in dark stores close to the customer, carry a narrow range, and decide city by city and store by store what goes on the shelf. A brand typically supplies stock rather than shipping orders, and is judged on whether it sells through quickly and stays in stock. The terms to negotiate are therefore a supplier’s terms: the margin the platform takes, payment days, who bears damaged and expired stock, listing and visibility charges, minimum marketing commitments and the fill rate you must maintain. Platforms that control demand can charge for it; Inc42 reported sellers paying 30 to 40 per cent on the large food-delivery platforms against about 5 per cent on ONDC buyer apps.
Quick commerce suits products bought often, on impulse or in a hurry, with a brand a buyer picks off a small screen in a second, and a supply chain that can replenish dark stores in several cities without stock-outs. It does not suit a considered purchase or a long catalogue. Start with one city and a handful of products, compute the contribution per unit after every term in the agreement, and expand only where the products sell through. The [ONDC lesson](/library/ondc-channel-or-distraction) covers the open-network alternative for local delivery.
The monthly marketplace P&L
On the tenth of each month, after the settlement reports arrive, build one profit and loss per marketplace and per top product. Lines: gross sales, returns and refused orders, net sales, product cost, every fee from the settlement report by type, ad spend, and contribution. Reconcile GST collected at source and income tax deducted against the marketplace’s statements. Mark every product whose contribution was negative two months running, and either reprice it, cut its ad spend to break-even, fix its returns or delist it. Check the rate card for changes; Amazon’s March 2026 cut is the reminder that fee structures change and that a product unprofitable last quarter may not be now. Then decide where next month’s ad budget goes, by contribution per rupee, not by sales.
Fee figures are as announced on the dates given and vary by category, weight and programme; read your own rate card before pricing. Nothing here is tax advice.