Your best seller might be your worst product

Every shop has a bestseller list, and every bestseller list lies a little. It counts what left the warehouse. It says nothing about what came back three weeks later, got unpacked, checked, restocked, and refunded.

We had a product like that. Top three by units for months. Then we counted the returns and it dropped to eighth.

So we rebuilt the list around a simpler question: what actually stayed sold?

Net units, not units

Analytics → Best sellers ranks by net units - ordered quantity minus what was returned or refunded. Cancelled orders and orders you’ve marked as ignored never enter the count. Every row also shows Times returned and, next to it, what share of sales that is.

That second number is the one worth staring at. Two products with 200 units sold are not the same business if one comes back 4% of the time and the other 22%. On the old list they sat next to each other looking identical.

There are three views:

• Products - the ranking most people want, aggregated across variants.
• Variants - the same list with SKUs, because “the black one in M” is usually the real story.
• By option - sales grouped by colour, size, or whatever options your products use.

The By option view surprised us most. Returns are rarely spread evenly across a product - they pile up in one size, or one colour that photographs differently than it looks. Grouping by option makes that visible in one screen instead of a spreadsheet.

The Returns tab

Next door, Analytics → Returns answers the other half: how much of what you sell comes back, over time.

• A daily returns chart, dated by when the return was recorded, with the period average and the same period before it for comparison.
• Top returned products and top returned variants, ranked, each row linking straight to the product page.

If the bar chart climbs after you change a supplier, a size chart, or a product photo, you find out in days rather than at the end of the quarter.

The counting rules we settled on

Return numbers are only useful if everyone agrees what counts. Ours:

• Exchanges are not returns. If a return ships replacement items, the customer still keeps something and the sale stands. Counting exchanges as returns would punish products with a size chart problem twice.
• Nothing is subtracted twice. A return that is also refunded takes one unit off the total, not two.
• Only recorded returns count as returns. A standalone refund with no items coming back reduces net sales, but it doesn’t appear in “times returned” - it isn’t a return, it’s a discount after the fact.
• Ignored orders stay ignored. Test orders and the ones you’ve excluded from metrics don’t quietly reappear here.

What we do with it

Three things, mostly. We stopped reordering raw materials for a product whose sales were half illusion. We fixed the description on the variant that dominated the returned list - the return rate halved. And when a product is genuinely good, the list now proves it, which makes the decision to put more money behind it much easier.

The bestseller list is one of those numbers everyone glances at and nobody questions. It’s worth questioning.