How much should you discount, and who should get it?

Set discount depth from your own margin, then check who actually redeems it. The break-even maths, where discounts leak, and how to test the strategy.

Author:Benson
DiscountingUpdated 12 October 202617 min read
Jump to a section

In short

Set depth from your own margin: a discount only pays if orders rise by at least the discount divided by your gross margin minus the discount, and that bar climbs fast. Then check who redeems it, since discounts often reach returning customers and codes spread beyond the audience you chose. Test offers against a holdout before going deeper.

It's the second week of October, and the offer for Black Friday, 27 November, is on the agenda. Marketing wants a quarter off sitewide: last year was a fifth off, competitors went deeper, and a bigger number feels safer.

Finance asks one question before anyone agrees: what did last year's discount actually buy?

You pull last year's sale-week orders and sort them by discount code. Most of the discounted orders came from customers who had bought before, some at full price a few weeks earlier. Further down are codes nobody in the room sent for that sale: a creator's code from the spring, the staff code, a welcome code on a customer's third order. A few orders carry two codes at once.

This guide covers the two decisions behind every offer: how deep to go, worked from your own margin, and who ends up with the discount, including shoppers you never meant to give it to. Each section links to a narrower guide where one goes deeper.

How much extra volume does a discount need to pay for itself?#

More than most offers bring. To keep gross profit level, a discount d at gross margin m needs unit sales to rise by d ÷ (m − d) over what you'd have sold at full price in the same weeks. The bar climbs steeply as the discount approaches the margin.

The discount comes off the price, but the product still costs what it cost. Business Victoria, the Victorian government's business service, publishes the table: 10% off at a 40% gross margin needs 33.3% more units just to keep gross profit level, 10% off at a 30% margin needs 50% more, and 15% off at a 30% margin needs twice the units. [Business Victoria (Victorian Government), Sep 2022] (opens the source)

Against operating profit the bar is higher, because price falls straight through to a thin bottom line.

18.7%

more volume is needed just to offset a 5% price cut and hold operating profit, for the average S&P 1500 company in McKinsey's 2003 analysis.

McKinsey Quarterly, Feb 2003(opens the source)

The same analysis found that a 1% price rise at stable volume lifts operating profit by 8%. [McKinsey Quarterly, Feb 2003] (opens the source) McKinsey sells pricing work and studied large companies, not a weekend code, so read it as the size of the bar, not a forecast.

That bar already makes the new orders pay for discounting the ones that would have happened at full price. Two things raise it further: some of the extra orders are only pulled forward from the weeks after the sale, so they aren't new, and gross margin after product cost flatters: take off the shipping you pay, payment fees and returns, and the margin you're really discounting against is lower. For a sense of range, US listed apparel companies averaged a 56.88% gross margin in January 2026 and specialty retailers 35.30%, in NYU Stern professor Aswath Damodaran's dataset. [Aswath Damodaran, NYU Stern, Jan 2026] (opens the source)

Commission comes off the same slice. Pay a coupon, cashback or creator partner on a discounted order and you fund both, and on an order that would have happened anyway both are pure cost. The commission rates guide covers setting each partner's rate with that in mind.

Who actually ends up with your discount?#

Mostly people who were going to buy anyway, plus some you never chose. An offer reaches three groups: shoppers it persuades, shoppers who would have paid full price, and shoppers outside your chosen audience who found the code somewhere else. Only the first group earns the uplift in the table above.

68%

of discounted orders were placed by returning customers, against 11% of full-price orders, across 176 Australian ecommerce brands over 12 months in ProfitPeak's data.

Klaviyo (author James Hurman), May 2026(opens the source)

ProfitPeak sells profit analytics, and the report its data fed was produced in partnership with Klaviyo, which sells email and SMS marketing, so treat it as one dataset, not a benchmark. In the same data, 88% of order volume went through discounted orders. [Klaviyo (author James Hurman), May 2026] (opens the source) In that data, the discount mostly went to people who already knew the brand.

Peak week shows the same appetite in a different dataset. In Signifyd's Cyber Five 2024 release, 35% of orders carried a discount code, with an average discount of 28.8%. [Signifyd, Dec 2024] (opens the source) Signifyd sells payment-risk and policy-abuse protection and doesn't say where the merchants behind that figure are, so read it as the shape of peak week.

Then there's the audience you didn't choose. Followers repost a creator's code, coupon sites list it, and coupon extensions try it at checkout for everyone who has one installed. Welcome codes are worth more to find: SimplyCodes, a coupon site and browser extension that earns a commission when shoppers buy through its links, says new-customer codes copied on its platform over 30 days carried a median 20% off, against 15% for its other percentage-off codes. [SimplyCodes (author Sean Fisher), Apr 2026] (opens the source) The first-order discount guide covers what Shopify's eligibility settings stop and what gets round them.

Codes also outlive the sale they were made for: once one reaches a coupon site, it keeps working for as long as the code itself does, which is why every code needs an end date.

The cost lands in more than one place. Finance sees the discount rate creep up with no campaign behind it, and a code that spread can carry a partner's commission as well as a discount. Which kind of code you give each channel decides how far it can travel; the shared, unique and single-use codes guide goes channel by channel.

What does a discount do to the orders that come after it?#

It depends on who got it. A deep discount can win first-time customers who come back, and teach established customers to wait for the next one.

The clearest evidence is three randomised field experiments with a US mail-order catalogue selling its own-brand durable goods, tracked for about two years by Eric Anderson and Duncan Simester. Deeper discounts increased later purchases by first-time customers and reduced them among established customers, which the authors link to buying ahead, drawing in buyers who value the products less, and growing sensitivity to deals. [Marketing Science (Anderson and Simester), Feb 2004] (opens the source) The same depth can be an acquisition cost for one group and a lasting price cut for another, so decide the audience before the depth.

Sales also borrow from the weeks after them. In ProfitPeak's data, revenue in the two to four weeks after a discount period fell by an average of 27% against each brand's normal baseline. [Klaviyo (author James Hurman), May 2026] (opens the source) Judge a sale on the whole month, not the sale week.

The message also does some of the work the discount gets credit for. Navdeep Sahni, Dan Zou and Pradeep Chintagunta analysed 70 randomised email experiments run by a large online ticket resale platform, comparing customers sent a discount offer with customers who weren't.

90%

of the extra spending from emailed discount offers came without the offer being redeemed, across 70 field experiments at an online ticket resale platform.

Management Science (Sahni, Zou and Chintagunta), Aug 2017(opens the source)

The authors read the offer as partly advertising. Counting redemptions credits the discount with the message's work, and can't say whether a smaller offer, or none, would have done as well. Only a group that didn't get the offer can.

Do you have to go deeper on Black Friday?#

No. Depth is your most expensive lever in the week your margin is thinnest, and at least one brand found going deeper didn't pay.

Adobe, which sells analytics to retailers, recorded US holiday discounts in 2025 peaking at 30.9% off listed price for electronics, 25.1% for apparel and 18.8% for furniture, [Adobe Digital Insights, Jan 2026] (opens the source) and expects up to 30% off in Cyber Week 2026. [Adobe Digital Insights, Sep 2026] (opens the source) In Klaviyo's data, overall Black Friday Cyber Monday discounts fell from 29.6% in 2024 to 26.2% in 2025, and Klaviyo says the brands offering the smallest discounts saw some of the healthiest sales growth. [Klaviyo, Dec 2025] (opens the source) Klaviyo sells segmentation it positions against blanket discounting and gives neither a sample size nor how it measured discounts, so take the direction, not the number.

Petite Plume, a pyjama and home brand, offered 20% off for Black Friday for years, went to 25% in 2024 and returned to 20% in 2025 because, its CMO told Modern Retail, demand "didn't move enough to justify the additional discount". [Modern Retail, Nov 2025] (opens the source)

Peak is also when discounts collide. On Shopify, discounts combine only where each one's Combinations settings allow it. Once they do, product discounts apply first, order discounts apply to the reduced subtotal, and two percentage order discounts are both calculated on the original subtotal: 10% and 20% off a $100 order leaves $70, not $72. A shopper can enter up to five product or order codes plus one shipping code on an order. [Shopify Help Center, Oct 2026] (opens the source) The discount combinations guide maps each rule to what it costs.

Whatever depth you choose, set it up early. Combinations, end dates and anything that touches your theme or tags need testing in ordinary weeks, before your code freeze stops site changes.

Won't cutting discounts lose your best customers?#

It might, so test a cut before you make it. The customers who wait for a code are often good customers, and finance and marketing will both say so.

The louder worry is that conversion falls without discounts.

It may, and gross profit per visitor can still rise, because the orders that remain carry full margin. Stopping every discount at once teaches you nothing you can separate from the season. Hold out instead: keep the offer for most eligible shoppers, withhold it from a random slice, and compare gross profit per visitor (revenue after discounts, minus product cost, divided by visitors) over whole weeks. If the slice without the offer earns as much, the discount was a subsidy; if less, you've measured what the offer is worth. The holdout guide covers sizing, stopping rules and reading the range.

What can you change with your own tools?#

Most of the levers are already in your platform. Each one narrows who gets a discount or how deep it goes, and each has a catch.

1. Target offers by who the shopper is. Shopify discounts can be limited to specific customers, customer segments or markets, and "Limit to one use per customer" tracks the buyer's email address or phone number. Keep the deepest offer for people who haven't bought yet. The catch: segment discounts only reach existing customer profiles, so a first-time buyer isn't in a "never ordered" segment until after their first order, and a merchant reported in January 2026 that a new email address gets round one use per customer.

2. Run fewer, shorter sales. Fewer events give customers less reason to wait. In the same report, brands running 11 or more promotional events a year saw profit margins fall 11%, though it doesn't say whether that comes from Klaviyo's global data or ProfitPeak's Australian brands. [Klaviyo (author James Hurman), May 2026] (opens the source) The catch: a sale that says it ends has to end. The CMA's guidance says a substantially similar offer appearing shortly after a genuine end date can still mislead.

3. Offer something other than money off. A free shipping threshold, a gift with purchase or a bundle aims to raise order size instead of cutting price, and a gift costs you its product cost, not its shelf price. The catch: free shipping isn't free to you. At one online retailer that tried many shipping-fee schedules, studied in 2006, free shipping and free-over-a-threshold offers brought extra sales, but the lost shipping revenue and the shoppers who didn't respond made them unprofitable. [Marketing Science (Lewis, Singh and Fay), Jan 2006] (opens the source) The free shipping threshold guide works out where to set the line and how spend-and-save tiers stack on it.

4. Give each channel its own kind of code. Unique or single-use codes for email and customer service, a named code per creator, and a shared code only where you mean it to be public. The catch: a creator can't read a unique code out in a video, and a unique code in one inbox doesn't stop an extension applying a different, shared code at checkout.

5. Set combinations on purpose and give every code an end date. The catch: locking combinations down also breaks stacks you want, such as a welcome code on a multi-buy, so test a cart that hits every discount before each sale.

6. Hold out a share of shoppers from each offer. Since 2 October 2026, Shopify's Rollouts let you send a discount to a share of online store traffic and compare it with your existing offers, with your eligibility rules still applied. The catch: the changelog describes reviewing conversion, and a deeper discount can win on conversion while losing on margin. Read gross profit per visitor from your order export, run whole weeks, and don't start a test during a sale.

How Benson fits a discount strategy#

Benson covers the part of a discount strategy your own tools can't see: what happens to your codes in sessions where a coupon extension runs, and what shoppers find when they go looking for a better code.

See where your codes have gone. The free code check shows which of your codes coupon extensions and voucher sites hold for your store, and how often they're used. Once the tag is installed, detection and observe mode run on every plan, Free included, showing which coupon extensions and AI shopping agents appear in your sessions.

Decide which codes work in extension sessions. On Stop and above, checkout code policies mean only codes on your allowlist work in coupon-extension sessions; anything else gets a friendly message instead of a discount. Allow your public sale codes and the codes of partners and creators you pay, leave staff, apology and other private codes off, and decide welcome codes one by one: when extensions account for most of a code's use, Benson suggests taking it off. Shoppers without an extension still get every code you gave them. The Checkout rules page lists the codes used most in extension sessions, and code watch shows whether each code that spreads is contained, replaced, or its replacement is spreading too. On Shopify, no app can change what an extension shows inside checkout, so the protection there works on codes; on your storefront pages, Block hides extension pop-ups by rule.

Sample store
A strict code policy at checkout: an extension tries a code that isn't on the allowlist, and the shopper sees a friendly message. Sample data.

Show your own offer first. On Own, your discount widget shows verified codes that apply in one tap, with stretch goals that tell the shopper how much more to spend for a bigger reward, so they see your offer instead of hunting for someone else's. Targeting decides who sees which offer: new or returning visitors, traffic source and campaign, referrer, basket value, device, country, schedule, and whether a coupon extension is present. It can ask for an email address or a social follow before revealing a code, and those codes never sit in its public configuration. That's the audience decision from earlier, made per session: a welcome offer for new visitors, a stretch goal for returning ones, nothing extra for shoppers already holding your sale email. You pay a flat plan price, not a commission on every order.

Own the moment: show your own verified offer before an extension shows one.

Test the offer, not just the code. On Scale, A/B tests compare two sets of codes, or the widget against no widget, on revenue per visitor, conversion, code use or order value, with a guardrail on discount rate that can stop a test that's giving too much away. No verdict appears before seven days and enough sessions and conversions in every group. On Stop and above, a holdout can sit behind Block or the strict code policy, and before a sale the readiness report lists sale codes extension users could apply.

See Benson on your store, or follow the own-the-moment playbook. The trial is 14 days of Own, no card.

Questions merchants ask

How much discount should I give?
Work out the break-even uplift first: the discount divided by your gross margin minus the discount. That's the rise in orders the offer has to create just to stand still. If it can't plausibly bring that many new orders, go shallower, offer something other than money off, or narrow who gets it.
Do discounts train customers to wait for a sale?
They can. In a catalogue field experiment, deeper discounts reduced later buying by established customers, while two others found they increased it among first-time customers. That's why who gets an offer matters as much as how deep it goes.
Should we match last year's Black Friday discount?
Not by default. One brand that went deeper found the extra demand didn't justify it and went back. Work out the break-even uplift for each depth, and test the cheaper option in the ordinary weeks before your code freeze.
Should I stop discounting altogether?
Not all at once. Withhold the offer from a random slice of eligible shoppers, compare gross profit per visitor over whole weeks, and cut where the slice without the offer earns as much.
How do I stop codes reaching people I didn't send them to?
Give each channel the right kind of code, set end dates and usage limits, and keep combinations off unless you've planned the stack. A shared code can still spread. With Benson's checkout code policies, on Stop and above, only codes on your allowlist work in coupon-extension sessions; on Shopify that runs through the Benson Shopify app.
Do I need to show a was price when I run a sale?
If you advertise a reduction, the reference price has to be genuine. In the EU, an announced reduction must show the lowest price of at least the previous 30 days. In the US and UK, the former price must have been genuinely offered for a meaningful, recent period.

Sources#

Keep the decision moving

Find out what's leaking. 14 days of Own, no card.

Start free trial