Spanish Business-Transfer Index, Issue 1 (August 2026): exposure, disclosure, seller concentration and asking prices in small-business transfer advertisements

A count of small-business transfer advertisements on one Spanish classifieds marketplace, published as an open dataset so that every figure can be quoted with the sample it rests on. Business-transfer advertisements on one Spanish classifieds marketplace. In the figures counted on 2 August 2026 the unit is the advertisement. In the asking-price figures, counted on 12 and 13 August 2026, advertisements judged to repeat one property are merged and counted once, so the unit there is a group of advertisements treated as one property. In neither case is the unit a completed sale. This dataset does not measure completed transactions, agreed prices, time to sale, and any business not advertised on that marketplace. What is in it — 72 measured rows, 9 columns Exposure: how long advertisements have been on the market. Median 146 days; first quartile 54 days; third quartile 440 days; ninetieth percentile 846 days (344 listings, counted 2 August 2026). 28 per cent are older than a year and the single oldest is 3,918 days; abandoned listings were not separated from live ones. Refresh: 96 per cent of advertisements look refreshed today or yesterday (344 listings). Disclosure, from 140 listings read line by line: asking price 90 per cent; monthly rent 41 per cent; floor area 36 per cent; years left on the lease 14 per cent; revenue 10 per cent; profit 1 per cent; employee count 0 per cent — not one of the 140. Sellers: 511 listings traced to 241 distinct sellers. 81 per cent come from professional accounts, 62 per cent from sellers holding more than one listing, and the ten largest sellers hold 43 per cent of the section; the single largest holds 105 listings. Asking prices for twelve named trades plus an "everything else" catch-all, thirteen rows in all, each carrying its own sample size: academies and training centres, bakeries, bars, clinics, dry cleaners and laundrettes, food shops and supermarkets, hair salons, money-transfer shops, nurseries, restaurants, tobacconists, vehicle workshops. Asking price per square metre (243 listings) and asking price expressed in years of the annual rent on the premises (54 listings). Three trades carry no price row, and the dataset records why. Pharmacies: 13 of 172 pharmacy listings state a price at all. Gyms: 20 of the 27 listings offering an actual gym state a price, below the 25 this index requires before comparing a price with a whole section. Hotels and guesthouses: this row records a decision, not a measurement: the portal keeps hotels and guesthouses in one section, and asking prices there run from 1,000 EUR to 120,000,000 EUR across 80 priced listings — too wide for a single band. The rows recording those decisions are part of the dataset rather than a silent omission. What this dataset does not measure "Time on the market" here is exposure — how long an advertisement has been up. It is not how long a sale takes. Datasets that measure completed sales, for example marketplace reports that publish a median days-on-market for sold businesses, measure a different quantity and are not directly comparable to this one. Relationship to the trade pages on the publisher's site The publisher also runs a separate page for each trade. Those pages and this dataset are built from the same reading of the marketplace, and for five of the thirteen priced trades — bakeries, dry cleaners and laundrettes, hair salons, money-transfer shops and nurseries — they report a different median for the same day. For the other eight the two agree. Both are classified counts rather than raw section totals: each trade page states which advertisements it kept and which it set aside, and the asking-price rows here rest on a selection of the same kind, described in limit 10. Why the five differ has not been established, and this dataset does not offer a reason for it. Which of the two is the better measure of a trade has not been determined either. An earlier version of this passage said that this dataset applied no such classification and counted the marketplace's sections as the platform publishes them, and called it the broader of the two measurements. Both statements were wrong. Limits, stated by the publisher 1. One marketplace. This is one Spanish classifieds platform, not the Spanish market. Portugal is not covered.2. Advertisements, not transactions. Asking prices are what sellers ask, not what anyone paid. Time on the market is exposure, not time to sale.3. Duplicates were treated differently in the two halves of this dataset. In the figures counted on 2 August 2026 duplicates were not removed: listings were counted, not premises, and one property advertised by two intermediaries enters the count twice. How often that happens has not been measured. In the asking-price figures, counted on 12 and 13 August 2026, republications of the same property were merged and counted once, advertisements from buyers looking to purchase were excluded, and prices too low to be asking prices were dropped; limit 10 says how far each of those steps can be checked. An earlier version of this notice stated the first half of this for the whole dataset. That statement was true of the 2 August figures and false of the asking-price figures. No measured value changed when it was corrected.4. Abandoned listings were not separated from live ones, so the upper end of the age distribution is inflated by an unknown amount.5. Exposure, refresh, disclosure and seller figures come from a single day (2 August 2026). Seasonality is not measured.6. Sectors come from the platform's own categories, not from an official classification. Three trades (pharmacies, gyms, hotels and guesthouses) have no price row because their priced samples are too small or their range too wide; they fall into 'everything else'. The rows recording that decision are part of the dataset.7. A price row is published only where at least 25 listings in the segment state a price. Segments below that threshold are folded into 'everything else' or recorded as a no-row decision.8. The food-shops section was not read in full. The portal claimed 746 listings; 327 were read, about 44 per cent. Of those, 206 were genuine sector offers and 164 stated a price, and the asking-price quartiles for that trade rest on those 164 — not on the 327 that were read.9. The median is a median. A statement such as 'half have been on sale longer than 146 days' says more than the data supports; what the data supports is that the median listing age is 146 days. The data does not distinguish listings whose age is exactly 146 days. See also limit 4: abandoned listings inflate the ages of the oldest listings.10. How the priced population was chosen, and how far that choice can be checked. For each trade the asking-price rows rest on advertisements offering a business of that trade, after republications of one property were merged and counted once, after advertisements from buyers looking to purchase were removed, and after prices too low to be asking prices were dropped. How far each step can be checked against the published figures differs by trade. Merging is load-bearing for seven of the thirteen priced trades, and the removal of buyers' advertisements for five, in the sense that without those steps the published figures cannot be obtained at all; for the remaining trades the figures come out the same either way, so they do not show whether the step was applied. The threshold below which a price was treated as a placeholder differs between trades, and the trades were read on different days. Taken together, the published figures fix one and only one selection of advertisements for six of the thirteen priced trades; for the other seven, several different selections yield the same figures. For hair salons and for nurseries the figures can be reproduced only if advertisements above a high price are also excluded, and no record of such a rule survives. The publisher holds the advertisement records behind these rows, each carrying the classification it was given at the time of counting. Licence and how to quote Published under CC BY 4.0: reuse and redistribution are permitted, in any medium and for any purpose, subject to the attribution and notice conditions in Section 3(a) of the licence — credit the source, keep the licence notice, link to the licence, and indicate any changes made. In addition we ask — but, under this licence, cannot require — that any figure quoted from this dataset carry the sample size printed beside it. A percentage without its sample size is an opinion wearing a number's clothes, and that is how figures like these get misused. Cite as: Provenance Diligence, Spanish Business-Transfer Index, Issue 1, August 2026, revised 22 August 2026 — https://doi.org/10.5281/zenodo.22057897 Versions and corrections Files in a published Zenodo record are immutable. Corrections are issued as a new version with its own DOI; the concept DOI always resolves to the latest version. A correction to the numbers in Issue 1, if one is needed, is published as a new Zenodo version of this same record. Issue 2 is a separate, later count, not a correction mechanism. The index is also published as a web page in three languages — English, Spanish and Portuguese — with the same Spain-only scope; the Spanish and Portuguese pages are translations, not separate country editions. See Related works. Revision history 2026-08-04: Issue 1 published. 2026-08-19: Issue 1 revised. 2026-08-21: Issue 1 revised: the pharmacies row carried a superseded denominator (157; the corrected count is 172); the gyms row carried a figure with no sample size (now 20 of 27); and the food-shops note conflated the listings read (327) with the sample the price quartiles rest on (164). No measured value changed. 2026-08-22: Issue 1 revised: the dataset described a method it had not used. It stated that duplicates were not removed, that republications were not merged and that advertisements from buyers looking to purchase were not removed. Applied literally, that method reproduces none of the thirteen asking-price segments. The asking-price figures rest on a classified population; the figures counted on 2 August 2026 do not. Limit 3 and the passage on the trade pages are rewritten, and a new limit 10 records how the priced population was chosen and what remains unknown about it. No measured value changed. Counted on 2026-08-02, 2026-08-12, 2026-08-13. Quarterly; Issue 2 repeats the same counts so that movement between issues can be read.