Did my sale make money?

How this works and its limits

How it works

The checker works out what you'd have sold without the sale, from your own report: the level of the four weeks before it, and the shape of your year and your week. What you sold, minus that, is what the sale added. The range is where the true figure probably lies, nine times in ten. It says "Real change" only when the difference is bigger than your sales usually move on their own.

For break-even, it uses the margin you type, the discount in your report, any price cut and any ad spend you add: a deeper discount earns less on each sale, and ads have to be paid for, so the sale needed more extra sales to pay for itself.

Its limits

• It can't tell your sale from anything else that happened at the same time. A busier season than last year, press coverage or a competitor running out of stock all look the same in your sales.
• With under two years of history, the shape of your year comes from one earlier year, so a season that's busier than last year for other reasons can look like a sale effect.
• Days in a row with no sales are taken as a closure. Orders held over from a closure can make a sale just after it look better than it was.
• Other sales in the period are left out of what's expected. We suggest them from your report's discounts: add any we missed, such as one run by lowering prices rather than with a discount.
• Small effects are hard to see when your daily sales jump around a lot. "Expected" means any effect was too small to tell apart from your usual ups and downs, not that the sale did nothing.
• Customers who bought during the sale instead of later make it look better than it was. That's why the result also shows the three weeks after it.
• Your margin is the one you type, and product costs are taken as unchanged for each pound of full-price sales.
• One margin for the whole sale: if it moved buyers towards lower-margin products, it'll look a little better than it was. The Stats Compass app uses each product's actual cost.
• Returns are taken off on the day Shopify records them (Net sales). Returns of sale orders that come back during the sale are taken off its sales; ones that come back later show in the three weeks after, not in its break-even, and ones that haven't come back by the end of your report aren't counted.
• Other campaign costs (design, influencers, free gifts) aren't included unless you add them to ad spend.
• Sales run by lowering prices don't show in Shopify's discounts; enter the average cut.

How it did on real data

We tested it on two years of a real UK online retailer's daily sales (the UCI Online Retail II dataset), on hundreds of dates with no sale, and with sales of a known size added in:

• On dates with no sale, it said "Real change" at most 4% of the time.
• A sale that lifted sales by 80% for one to three weeks was picked out about two times in three, and one that lifted them by 50% about one time in three. Smaller lifts were rarely picked out.
• The "likely between" range held the true effect about nine times in ten.

That retailer sold mostly to wholesalers, in fewer and larger orders than a typical Shopify store, so its daily sales jump around more than most.