data study
We measured 6,109 live sale prices. The discounts come off a menu.
Price & Thread editorial · Jul 20, 2026 · every figure measured from our own tracking data, snapshot July 20, 2026 · methodology
A discount is supposed to be a fact about the past: this cost more, now it costs less. Measure 6,109 of them at once — every live clothing markdown across the 19 U.S. stores in our July 20, 2026 snapshot, each one carrying the store's own compare-at claim — and a different picture appears. Discounts don't look like history. They look like a menu.
Six numbers do most of the work
If markdowns reflected item-level reality — this jacket over-ordered, that colorway dead — claimed depths would scatter. They don't. More than two-thirds of the sale section of the American internet is set to one of six clean percentages:
| Claimed depth | Live offers | Share of all 6,109 |
|---|---|---|
| 30% off | 870 | 14.2% |
| 40% off | 830 | 13.6% |
| 50% off | 794 | 13.0% |
| 20% off | 729 | 11.9% |
| 70% off | 582 | 9.5% |
| 60% off | 410 | 6.7% |
Together: 4,215 offers — 69% of everything on sale — round to 20, 30, 40, 50, 60, or 70 percent off. Hold the ruler tighter and the pattern survives: 65.5% of all 6,109 claimed depths land within a tenth of a percentage point of an exact multiple of 5%. That is not the fingerprint of six thousand individual pricing decisions. It is the fingerprint of a spreadsheet: someone selects a collection, types a percentage, and a thousand “was” prices are born at once.
The sticker prices are built for the math
The compare-at prices themselves have a telltale shape: 82.6% end in .00 and another 16.0% end in .99 — between them, 98.6% of every former price on the menu. Clean round numbers divide into clean percentages and produce sale prices like $106.80 (exactly 40% off $178) and $13.80 (exactly 70% off $46). When you see a sale price with odd cents, you're usually looking at the output of a percent rule applied to a round sticker — the discount was designed, not discovered.
How deep the claims run
The median live markdown claims 40% off. A full third of them — 2,057 offers, 33.7% — claim you're saving half or more. One in nine (672 offers) carries a compare-at price at least triple the asking price. And almost nothing is modest: just 24 offers out of 6,109 — 0.4% — claim less than 10% off. In claimed discounts, as in fishing stories, small catches go unreported. The deepest claim in this snapshot: 84.5% off, at Kith.
The same “40% OFF” is a different fact at different stores
The most useful thing we measure is each store's discount culture: the median and 90th-percentile depth across its entire sale section. Eleven of our stores had sale sections big enough to measure reliably (our floor is 200 sale SKUs) on July 20, 2026:
| Store | Median markdown | 90th percentile | Sale SKUs measured |
|---|---|---|---|
| Fashion Nova | 26% | 50% | 6,340 |
| Gymshark | 70% | 70% | 4,528 |
| Taylor Stitch | 39% | 50% | 2,617 |
| Marine Layer | 30% | 50% | 1,793 |
| Kith | 55% | 70% | 1,684 |
| True Classic | 33% | 51% | 710 |
| Alo Yoga | 30% | 31% | 442 |
| Outdoor Voices | 40% | 70% | 426 |
| Mack Weldon | 7% | 50% | 411 |
| Fresh Clean Threads | 24% | 49% | 390 |
| American Giant | 37% | 60% | 368 |
Read that table twice and “% OFF” stops being a number and becomes a dialect. At Alo Yoga, the sale section barely moves — median 30%, and even the 90th percentile sits at 31% — so a 30% cut there is the store going about as deep as it goes. At Gymshark, 70% off is simultaneously the median and the 90th percentile: the whole sale section sits at one depth, which makes “70% OFF” roughly as newsworthy as a Tuesday. Mack Weldon's median markdown is a token 7%, but its tail reaches 50% — a store where the percent tells you almost nothing until it's large.
What to do with this as a shopper
- Judge the percent against the store, not against zero. A 40% cut at a store whose median is 39% (Taylor Stitch) is the store behaving normally. The same 40% at Alo Yoga would be an outlier worth a look.
- At flat-depth stores, shop the dollar price. Where everything is “70% off,” the percentage carries no information — compare final prices instead.
- Treat triple-priced compare-ats as decoration. When the sticker is 3× the ask — 11% of everything we track — the “was” price is doing theater, whatever the garment is doing.
How this feeds our labels
These same snapshots are one of the evidence sources behind every badge on this site — with a rule that matters: a store's pattern corroborates only claims that fit the pattern. A 40% claim at a store that routinely cuts 40% earns its corroboration; a 75% claim at that same store corroborates nothing. The full ladder is at how we verify, and the narrative version is how we verify a sale.
Method notes
Snapshot taken July 20, 2026 from our own catalog: 6,109 live markdowns (a “sale” here means the store lists a compare-at price above the asking price) across 19 U.S. direct-to-consumer stores, sizes and duplicates deduplicated to one offer per product per store. Claimed depths are computed from the retailers' own compare-at prices — this study measures what stores claim, which is exactly why we don't take the claims at face value elsewhere on the site. Our store mix skews toward Shopify-based DTC brands, not department stores, and these figures describe one day — counts will drift as inventory changes, and we'll re-run the study as our history deepens. Questions about the data: ask — journalists can have the underlying numbers.