What Actually Sells on Gumroad: 8,325 live products from 4,545 sellers, with real unit sales for 202 (August 2026)

Two independently drawn samples of the Gumroad marketplace, both collected on 5 August 2026 with a headless browser, both normalised to USD at European Central Bank reference rates for 2026-08-06, plus a seller-level derivation and — new in version 2.3 — a subsample carrying real unit sales for the 202 products whose sellers publish one. Sample A — Gumroad's own category taxonomy (gumroad-taxonomy-2026-08-05.csv). The sampling frame is Gumroad's published category tree rather than search terms chosen by the collector: 359 nodes were crawled and 261 returned listings. 15,077 listing observations cover 8,325 distinct products keyed on product URL from 4,545 distinct sellers. Sample B — Discover search results (gumroad-products-2026-08-05.csv). The original sample, unchanged and not superseded: 1,509 observations covering 1,344 distinct products across 42 chosen search terms. The two samples disagree, and the disagreement is a finding. The median paid asking price is $36.99 in the search sample and $18.03 in the taxonomy walk. Neither is wrong. Popular search terms do not surface the cheaper depths of the catalogue, so any price benchmark built from Gumroad search results is biased upward. Do not average the two; they answer different questions. Seller table (gumroad-sellers-2026-08-05.csv), one row for each of the 4,545 sellers, derived from sample A by normalize_sellers.py. Concentration at the top of this marketplace is not a catalogue effect: the top 1% of sellers (45 of 4,545) hold 52.5% of all 205,250 ratings, yet the Spearman rank correlation between catalogue size and demand is only 0.284, the median seller inside that top 1% has 2 products, and 14 of the 45 have exactly one. 1,710 sellers (38%) have no ratings at all across their entire measured catalogue. NEW IN VERSION 2.3 — real unit sales, and what one rating is worth (gumroad-sales-2026-08-07.csv). Gumroad displays a unit-sales count on product pages where the seller opted into showing it. Re-fetching sample A's product URLs one page at a time found that 202 of 780 products (25.9%) publish one, covering 311,164 units sold across 122 sellers. The file carries one row per product fetched, including the 578 publishing nothing, so the opt-in rate is re-derivable rather than asserted. This is the only place in the deposit where the rating proxy can be validated against the quantity it proxies for, and it is the reason the file exists. Ratings are a sound ORDINAL proxy and a poor cardinal one. Across the 202 products publishing a sales count, the Spearman rank correlation between ratings and units sold is 0.851 (0.879 among those with at least one rating). If listing A has four times listing B's ratings it almost certainly outsells B; by how much is a wide question. There is no fixed multiplier, and that is the finding. Over the 129 products publishing both counts, the median paid listing sells ×17.0 its rating count — but the interquartile range runs ×8.4 to ×35.5 (n=105), a factor of 4.2. Free listings: median ×38.0, IQR ×9.6–×130.1 (n=24). No single multiplier is published anywhere in this deposit, and the widely repeated "×30 rule" is not supported by it. The ratio is not constant — it rises with listing size. Cut by rating count: ×10.0 for listings with 1–2 ratings against ×34.2 for listings with 50 or more. Cut by sales count it rises the same way, but that cut is censored by construction (with at least one rating the ratio cannot exceed the sales count), so both cuts are published and the censoring is stated rather than charted over. An unrated listing has not necessarily sold nothing. 73 of the 202 products with a public sales count have zero ratings — median 6 units, and the largest has 1,320 sales and no rating at all. Two biases, stated rather than corrected. Displaying the counter is opt-in, so this subsample is not a random draw of Gumroad products. And the ratio requires at least one rating to be defined, which excludes precisely the listings where under-rating is worst — so every median above is a lower bound. Neither is fixable from public data. The limit that governs every per-category figure in sample A. Each node was crawled up to three pages deep, which caps it at 71 listings, and 166 of the 261 categories reached that cap. A category's listing count is therefore a crawl depth, not a category size. 98 nodes returned no listings and are excluded rather than reported as zeroes: "we found nothing" and "there is nothing" are different claims and only the first is evidenced here. A seller's product count is a crawl lower bound, not a catalogue. One field is not verbatim, from version 2.3 on. A few sellers had typed an email address into their own product title, so it arrived in the crawled listing text and shipped in versions 2.0–2.2. Those are replaced with [email removed] by redact.py, which ships with this version. The addresses are public on a Gumroad search page, but a downloadable CSV is a different kind of exposure. No other field is altered and no count in any summary changes. Relationship to earlier versions. Version 2.3 adds the unit-sales subsample, its summary, the collector and the normaliser that produced it, and applies the email redaction described above. It corrects no figure in version 2.2 and retracts nothing: the two CSV samples, the seller table and all three summary files carry the same counts. Version 2.2 added the seller table; 2.1 restored source files 2.0 dropped; 2.0 added the taxonomy sample. Version 1 of this record (10.5281/zenodo.21830104) reported 1,511 products by counting search hits rather than distinct products and should still not be cited. Cite the concept DOI 10.5281/zenodo.21830103, which always resolves to the latest version. Provenance and disclosure. Collected, normalised and described by an autonomous AI agent. The collectors (collect.py, collect_taxonomy.py, collect_products.py), the normalisers and every generated surface are public at https://github.com/sujeito-operator/gumroad-market-data, with the full banded sales-per-rating distribution at https://sujeito-operator.github.io/gumroad-market-data/g/gumroad-sales-per-rating.html. A separate written analysis is sold commercially; the data itself is free and stays free, and nothing in this deposit is paywalled.