Method and limitations
Source. Weekly snapshots of the Shopify storefront population, June 2020 – July 2026. Base week 2020-W26, comparison weeks the nearest available to each June, final week 2026-W31.
Eligibility. Active storefronts with ≥10 products, a known price and a known category. The 10-product floor excludes stores whose week-to-week movement is indistinguishable from noise.
Matched cohort. Every change figure is computed on domains present in both the base and comparison week. Composition is held constant by construction.
Medians, not means. Price and SKU distributions are heavily right-skewed; a mean describes no actual store.
Limitations — state these plainly
- Prices are modelled, derived from storefront data rather than observed transactions. They are best read as relative movement, not absolute price levels.
- Survivor-only. The matched cohort answers "did merchants raise prices", not "what happened to the market". Stores that failed are counted in survivorship and nowhere else.
- "Survival" means continued presence in the eligible tracked population — not strictly that a business still exists. We re-checked a random sample of 2,984 storefronts the study counts as not surviving: 94.6% have left the tracked population altogether, but about 5% are still trading in some form — 4.1% remain on Shopify while no longer meeting the eligibility bar (typically having shrunk below ten products), 0.7% are live but password-protected, and 0.6% migrated to another platform (WooCommerce, Wix, BigCommerce) and therefore leave a Shopify-only cohort by construction. Adjusting for all of them moves the headline survival rate from 25.7% to roughly 30%. We report the stricter figure because it is the one the method actually measures, but "74% of 2020's storefronts went out of business" would overstate it — a meaningful minority shrank, closed their doors to the public, or moved house.
- No revenue weighting before 2024. All figures are unweighted store medians, so a $50M brand counts the same as a $50k one.
- Categories are self-reported primary classifications; multi-category merchants are counted once.
- Assortment counts products, not depth. A 500-SKU store may be one product in 500 variants.
- Assortment counts listed products. Verified against our own crawl: the median store carries 7.0% currently-unavailable and 1.6% long-abandoned listings, so listed and live assortment are close for a typical merchant — but we can measure that share only for 2026, and it rises with catalog size. Growth measured on listed products is +33.9%; on products excluding long-abandoned listings, +24.6%.
- Category is measured at a point in time. 10,137 surviving merchants (5.7%) changed primary category over the period and therefore appear in the cohort totals but in no category row.