पाठशाला Pathshala · हिसाब Hisāb, Unit economics · Lesson 16 · Build
Discounts, cashbacks and the margin you give away
A discount is taken from contribution, not from price, so a small one costs a large share of what each order leaves. Price every coupon and set the rules that make a discount an investment.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
A founder who would never cut the price of a product by a fifth will run a 20 per cent festive code without a second thought. They are the same decision. The difference is that the price cut is visible on the product page and the code is buried in the marketing calendar, where nobody adds up what it cost.
This lesson adds it up. It shows where a discount comes from in the arithmetic of an order, gives the formula for the extra volume any discount must buy, works one festive sale through, prices cashback and free delivery honestly, and sets the rules under which a discount earns its keep.
A discount comes out of contribution
A discount reduces revenue on an order and leaves every cost of that order where it was. The goods cost the same. The courier charges the same. The gateway fee falls a little and nothing else moves. So the whole discount is taken from [contribution](/library/contribution-margin-first-number-to-know), the money the order leaves after its own costs, and contribution is far smaller than price.
Take an order at ₹1,500 ex-GST that leaves ₹600, a 40 per cent contribution margin. A 10 per cent discount is ₹150, a quarter of the contribution. A 20 per cent discount is ₹300, half of it. A 30 per cent discount is ₹450, three quarters. The discount reads as a modest share of price and lands as a large share of profit. Michael Marn, Eric Roegner and Craig Zawada made the same point about price in The Power of Pricing: in their analysis a 1 per cent rise in price, with volume steady, lifts operating profit by 8 per cent, and a 1 per cent fall cuts it by the same. A discount is a price fall with a coupon code.
The volume a discount has to buy
Because each discounted order leaves less, the offer must bring more orders just to leave the month where it was. The break-even uplift is the discount divided by the margin less the discount. At a 40 per cent margin a 10 per cent discount needs 33 per cent more orders, a 20 per cent discount needs 100 per cent more, and a 30 per cent discount needs 300 per cent more. At a 30 per cent margin the 20 per cent discount needs 200 per cent more orders, and a 30 per cent discount can never break even, because every order leaves nothing.
That is the arithmetic before the two corrections that make it worse. First, many of the discounted orders would have happened at full price; those customers get the discount and the company gets nothing for it. Second, some orders are not new but pulled forward from the weeks after the sale, so the uplift in the sale week is followed by a dip. Measure uplift against the same weeks without the offer, and look at the four weeks after as well.
A festive sale, worked
A Lucknow brand selling chikankari kurtas online averages ₹1,500 an order ex-GST and keeps 40 per cent as contribution. In an ordinary month it takes 2,000 orders and ₹12 lakh of contribution. For Diwali it runs 20 per cent off sitewide and takes 3,000 orders, half as many again. The team calls it the best month of the year.
Contribution says otherwise. Each order now leaves ₹300, so 3,000 orders leave ₹9 lakh, ₹3 lakh less than an ordinary month. The company gave away ₹9 lakh of discount, ₹300 on every one of the 3,000 orders, to win 1,000 orders it would not otherwise have had: ₹900 of discount for each extra order, against the ₹600 a full-price order leaves. Revenue rose by a fifth. The business went backwards by a quarter.
The figure opens on the Lucknow sale. Drag the uplift to 100 per cent to find the point at which the offer only stands still. Then cut the discount to 10 per cent and see how much less volume the offer needs. Set the margin to your own and read the break-even uplift for the codes you plan to run this quarter, before they run.
Cashback, free delivery and bank offers
The label changes and the arithmetic does not. Cashback credited to a wallet costs nothing on the day and the full amount when it is redeemed, which is usually on a later order at a moment the company did not choose. Count it at the share actually redeemed, measured from your own data over at least two quarters, and charge it against the order that earned it. Free delivery is a discount equal to the courier cost, paid on every order whether or not the customer needed persuading. Bank and wallet offers are discounts too, at whatever share the company funds; read the agreement and book only the company’s share, all of it.
Listed Indian platforms define the line plainly. Swiggy’s Q4 FY25 shareholder letter computes contribution for each business after subtracting platform-funded discounts, defines net order value as gross order value less discounts whether platform- or partner-funded, and names elevated customer incentives, delivery-fee discounts among them, as a pressure on quick-commerce contribution that quarter. A seed-stage company should report the same way: discounts above the contribution line, by type, where the board can see them.
When a discount is an investment
Some discounts pay. They share three features. They go to people who were not going to buy: a first-order offer to someone who has never bought, not a sitewide code that every loyal customer also uses. They are repaid by later full-price orders inside a window the company’s cash can carry, which makes the discount a [customer acquisition cost](/library/cac-ltv-and-payback-the-three-numbers) and puts it in the same ledger as advertising. And their effect is measured against a holdout: a random tenth of eligible customers who do not see the offer, compared with those who do.
Write the rules down before the next sale. Every offer names its target customer, its cost per order, its break-even uplift and the holdout that will measure it. A first-order discount is capped at the repeat contribution a new customer leaves within ninety days. Sitewide discounts are limited to a set number of days a year, decided in the annual plan. Any offer whose holdout shows less uplift than its break-even is not run again in that form.
There is a cost the ledger does not show. Eric Anderson and Duncan Simester, writing in Mind Your Pricing Cues, found that customers know remarkably little about what prices should be and lean on the seller’s signals, sale signs among them, to judge whether a price is good; used inappropriately, they warn, those cues breach customers’ trust and reduce brand equity. A brand that is always on sale has told its customers what its real price is, and it is not the one on the label.
A discount is a price cut paid for out of contribution. It is worth running only when the extra orders it buys are worth more than the margin it gives to the orders that were coming anyway.
Every leak in the price, on one page
The McKinsey authors called the tool the pocket price waterfall: start from the list price and subtract every deduction between it and the money the company actually keeps from a transaction. In their example of a lighting supplier, once prompt-payment discounts, co-operative advertising, freight and other items that never appear on an invoice were counted, the average pocket price came to roughly half the standard list price. They also found companies could often capture an extra 1 per cent or more of realised price simply by managing those leaks.
Build the same waterfall for an online brand: list price, sitewide discount, coupon codes, cashback redeemed, free delivery, bank-offer share, marketplace-funded discounts recharged to you, and returns refunded. The bar at the end is what each order really brought in. Most founders who draw it for the first time find a step they had not counted.
The monthly discount ledger
On the fourth working day of each month list every offer that ran: code or mechanic, days live, orders that used it, discount given in rupees, uplift against comparable weeks, the four-week dip after it if any, and the holdout result where there was one. Total the discount given across all offers and divide by revenue, so the month has one discount rate that the board sees beside gross and contribution margin.
Then decide three things and write them under the ledger. Which offers ran above their break-even and may run again. Which ran below it and are retired or redesigned. And what the discount rate is allowed to be next month. Keep the pocket price waterfall on the same page and redraw it each quarter. A company that reviews its discounts monthly gives away less of its margin by accident, and when it does give some away, it knows what it bought.
Nothing here is legal, tax or investment advice. The Lucknow brand is illustrative; check consumer-protection rules on how offers and original prices may be displayed before you run one.
Sources
- Michael V. Marn, Eric V. Roegner and Craig C. Zawada, The Power of Pricing, McKinsey Quarterly, February 2003 — A 1 per cent price rise at steady volume lifts operating profit 8 per cent; the pocket price waterfall; pocket price near half of list in the lighting example.
- Swiggy Limited, Q4 FY2025 Shareholder Letter, May 2025 — Contribution after platform-funded discounts; net order value net of platform and partner discounts; customer incentives as a margin pressure.
- Eric T. Anderson and Duncan Simester, Mind Your Pricing Cues, Harvard Business Review, September 2003 — Price cues used inappropriately can breach customers’ trust and reduce brand equity.