The zip-code discount that beats a sitewide sale

Why a sitewide sale spends most of its budget on nothing

The sitewide sale is the default because it is easy to set up and it always produces a number that looks like success. Revenue goes up during the sale. Everybody agrees it worked.

Here is the same week with the baseline removed.

Assume 900 orders a month, an average order value of Rs 1,800, a 10% discount and an 8% incremental lift. The discount costs Rs 162,000 and produces 67 incremental orders, which is Rs 2,430 for each order that would not otherwise have happened.

The other 833 orders got a 10% discount for buying something they were going to buy anyway. That is 92.6% of the budget spent on nothing.

Nobody is shocked by this in the abstract and almost nobody prices it. The number is invisible in every dashboard, because dashboards report revenue during the sale rather than revenue attributable to the sale.

Everything that follows is an attempt to move that 92.6%.

Gating a discount does not save money on its own

The obvious fix is to stop giving the discount to everyone. Gate it to your best postcodes and stop subsidising the rest of the map.

The arithmetic does not cooperate.

Gate the identical 10% discount to the top 20 postcodes, which hold 371 of the 900 orders. Spend falls to Rs 66,780. Incremental orders fall to 27.

Cost per incremental order is Rs 2,430. Exactly what it was sitewide.

Gating shrinks the subsidy base and the lift in the same proportion, so the ratio between them does not move. You have spent less and bought proportionally less, which is a smaller sale rather than a better one.

This is the version most brands actually implement when they try geo-targeting, and it is why the conclusion is so often that geo-targeting does not work. The mechanic was never the problem. Holding the discount rate constant was.

The same budget, concentrated: what depth it buys

Hold the budget constant instead of the rate. The full Rs 162,000 gets spent, just into fewer postcodes at a deeper discount.

Zone	Orders per month	Budget-equivalent depth
Sitewide, 320 postcodes	900	10.0%
Top 100 postcodes	641	14.0%
Top 50 postcodes	514	17.5%
Top 20 postcodes	371	24.2%
Top 10 postcodes	281	32.0%
Top 5 postcodes	205	44.0%

Modelled on Rs 1,800 average order value and a Zipf distribution with exponent 0.8 across 320 postcodes. Substitute your own concentration curve before using these depths.

The top 20 row is the interesting one. The same money that buys a forgettable 10% everywhere buys 24% in the postcodes that already produce 41% of your orders.

Twenty-four percent is an offer people mention to each other. Ten percent is not, and that difference is the entire mechanism this article is about.

The bottom two rows are a warning. Concentrate too far and the implied depth passes 40%, which most D2C gross margins cannot absorb for a week regardless of what it does to conversion.

The lift the concentrated version has to clear

Concentration is not free and the break-even is calculable, which makes this a decision rather than a belief.

The sitewide version produces 67 incremental orders. The concentrated version has to match that from a smaller base of 344 baseline orders in the top 20 postcodes.

That requires a 19.5% conversion lift, against the 8% assumed sitewide. Roughly two and a half times the lift, from roughly two and a half times the discount depth.

So the honest position is that a linear response makes this a coin flip. If 24% off produces exactly 2.4 times the lift of 10% off, the two strategies tie and you have added operational complexity for nothing.

The concentrated version wins when the response is better than linear. Two things push it there.

The first is urgency. A discount that is available to a specific zone for a specific window cannot be planned around, and a shopper who cannot wait for the next sale buys now.

The second is density, which is the part that does not show up in any discount model.

Why density helps and reach does not

A shopper who likes an offer forwards it. Where that forward lands is the difference between the two strategies.

Forwards are geographically short. The WhatsApp message goes to a sister in the same city, a colleague in the same office, the building group, the school parents’ thread. Most recipients live within a few kilometres of the sender.

In a national campaign, that locality is wasted, because the offer is equally available everywhere and forwarding it changes nothing for the recipient. The message is information, not an opportunity.

In a zone-gated campaign, the same forward lands on people who are inside the gate and can use the offer today. The same message does more work because of where it lands, not because of what it says.

That effect compounds when referral is structural rather than incidental.

SB&R is a Shopify app for chained referral rewards. Every referral link belongs to someone who has already bought. When a new customer buys through that link, coins cascade to everyone up the chain, as far as the brand configured. Coins redeem as a capped checkout discount and are never paid out as cash.

Coins that outlive the window matter here. A recipient who buys during the live window and earns a balance has a reason to return after their zone goes dark, which converts a one week promotion into a retention mechanic.

Running the offer itself needs a gate the shopper cannot edit.

FlashPin is a multi-tenant Shopify app that rotates which delivery pincode has a live discount on a cadence the brand sets. Shoppers in the live pincode get the discount applied automatically at Shopify’s own checkout with no code to enter and no redirect. Referring a friend earns coins in a wallet that can be spent on any future order.

A code would defeat the entire design, because a 24% code posted in one group chat is national within a day. Gating on Shopify’s own delivery address field rather than a shopper-writable cart attribute is what keeps the concentration real.

How to run the test without fooling yourself

Pick a postcode from ranks four to eight in your order map, not the top one. Your densest zone buys regardless and will flatter the result.

Hold out a comparable postcode. Without one you cannot separate the offer from the month, and the month is usually doing more than you think.

Compare the live zone against its own previous four weeks. Never against the national average and never against a different postcode, because cross-zone comparison imports every difference between the zones into the read.

Measure cost per incremental order, not revenue during the window. Revenue during the window will look good in every scenario including the ones that lost money.

Watch return to origin in the live zone. A deep discount pulls in low intent cash on delivery orders, and a 24% offer that lifts orders 20% while lifting RTO from 8% to 18% has destroyed value while showing a positive headline.

Run at least three cycles. The first window in any zone is mostly discovery, and the useful numbers start once residents have seen the offer move at least once. The mechanics of the rotation itself are covered in a hyperlocal flash sale without a physical store, and the concentration curve you need first is in why your best customers sit in six postcodes.

When the sitewide sale is the right call

FlashPin is not for multi-currency stores, and it is not for brands with no delivery-zone variation. If delivery is identical across your map, concentration buys you depth without buying you any of the urgency or density that was supposed to justify it.

Clearance is the clearest case for going sitewide. If the goal is to move stock rather than to buy incremental customers, reach is the point and concentration works against you directly.

Thin maps are the second case. A brand doing 60 orders a month has no postcode with a readable baseline, and a concentrated test there produces a number that is entirely noise.

Category matters too. A considered purchase with a three month cycle will not respond to a seven day window in any zone, because the buyer’s timeline is set by something other than your calendar.

The limitation worth sitting with is that the whole argument rests on one unmeasured quantity: whether your response to discount depth is better than linear. Everything else here is arithmetic, but that number is empirical and it differs by brand, by category and by season.

Run the three cycle test before you move a festive budget onto this, and treat the model as a way to size the bet rather than as evidence it will pay. The underlying case for reading delivery data as a channel is in the marketing channel nobody has touched.