A sitewide discount is the most expensive way to increase orders, because it pays every customer, including the large group who had already decided to buy. The incremental orders it creates have to cover the margin given away on all the rest, and that arithmetic is much harder than it looks from the revenue chart.
The revenue chart is the problem. A sale week looks wonderful in revenue and rarely gets examined in gross profit. The orders arrive, the graph goes up, and nobody separates the customers who were persuaded from the customers who were queuing anyway.
We are not against discounting. We are against discounting as a reflex, applied to everything, on a schedule, with no test of whether a narrower offer would have done the same job for less. Almost always, one would have.
The useful question before any sale is not what the discount should be. It is which specific group of people we are trying to move, and whether they need money to move.
Work out what the sale actually costs
Run this before the next one and the conversation changes permanently.
Take your normal weekly order count and revenue. Take the sale week's order count and revenue. The extra orders are the ones the sale created, roughly, before seasonality is accounted for. Now calculate the discount given on all orders in the sale week, including those you would have received anyway.
A labelled example. Say a normal week is 200 orders at $80, so $16,000 revenue and, at a gross margin of five in ten, $8,000 gross profit. Sale week runs at a fifth off. Orders rise to 300 at $64, so $19,200 revenue. Gross profit per order is now $24 instead of $40, so gross profit is $7,200.
Revenue rose by $3,200 and gross profit fell by $800. That is the shape of the trap. The made-up numbers do not matter. The method does, and your own version of it takes twenty minutes with an export.
There are sales where the sums work, especially when you need cash or the stock is costing you money to hold. Just run the sums first.
The question is not what the discount should be. It is whether the people you want need money to move.
The habit is more expensive than the margin
The direct cost is one thing. The trained behaviour is worse, because it does not appear in the report on the sale.
Customers learn schedules quickly. If a store discounts every month, its regulars stop buying between discounts, and full-price weeks get quieter every quarter. The sale then looks increasingly successful, because it is capturing demand that the sale itself displaced.
The way to see it is to plot revenue for the weeks before and after each promotion, not just during. If the two weeks either side dip and the dip is deepening over the year, the promotion is moving revenue in time rather than creating it.
The same habit changes who your customers are. Repeated discounting recruits people who buy on price, and they return only when there is another discount. Your repeat rate can rise while the value of a repeat customer falls, which is one of the more confusing things to see in a dashboard.
Brands that sell on quality lose something else as well. A price that is never the real price teaches people to wait, and waiting is the opposite of the urgency the sale was meant to create.
Narrower offers do the same job for less
Almost every reason to run a sitewide sale has a cheaper version that targets only the people you actually need to move.
If the goal is clearing stock, discount the stock that needs clearing. A sitewide sale to shift a slow line pays discount on the products that were selling fine.
If the goal is a bigger basket, use a threshold or a bundle. Both raise the order rather than lowering the price, and both are paid only when the customer does the thing you wanted.
If the goal is winning back lapsed customers, send the offer to lapsed customers. That is a segment, not a banner, and it costs you nothing on the people still buying at full price.
If the goal is a first order from a new visitor, put the offer where only new visitors see it, and accept that some regulars will find it. That is a smaller leak than a public sale.
If the goal is a specific revenue number by a date, a shorter sitewide event with a real reason is defensible. Once a year, tied to something, with an end. Not a rolling condition.
What to do about the season when everyone discounts
There is one period where the pressure to run a sitewide sale is at its highest, and refusing outright is not always the right call. The question is what shape it takes.
Our preference is a deeper offer on a narrow selection rather than a shallow offer on everything. It gives shoppers a reason to look, gives you a headline number to advertise, and protects the margin on the rest of the catalogue.
Bundles work particularly well in that period, because gift buyers want the decision made for them. A set at a price that is fair but not discounted below its parts outperforms a percentage off, and it raises the order value while it does so.
Early access for your email list is worth more than an extra few points of discount. It rewards the customers you already have, generates orders before the expensive advertising period, and costs nothing.
And whatever you run, prepare the store for the volume rather than only the offer. A sale that arrives at a slow product page and a leaking checkout converts worse than a quiet week, which is a discount paid twice.
How we would decide, next time
Four questions, in order, before any discount is approved.
Who are we trying to move, specifically. If the answer is everyone, the answer is not thought through yet.
Do they need money to move, or do they need information. A large share of hesitations are about delivery, fit, returns or a question the page did not answer, and no discount fixes any of those. It just pays people to overlook them.
What is the cheapest instrument that reaches only that group. A segment, a threshold, a bundle, a product-level markdown, early access, or a public sale in that order of preference.
What will we measure, and against what. Gross profit for the period against the same period without the promotion, plus the two weeks after, so displaced demand shows up.
Run that sequence a few times and you will discover most planned sitewide sales turn into something smaller and better. The ones that survive the questions are usually worth running, and you will be able to say why.