The market layer: what 344 listings tell you about the one you are looking at

You are looking at one advertisement. You cannot see the other three hundred and forty-three.

That sentence is the whole reason this part of the report exists. Every judgement a buyer makes in front of a listing — that looks expensive, that has just come up, they're hiding the takings — is a comparison against a market the buyer has imagined. The imagined market is usually wrong, and it is wrong in ways that consistently favour the seller.

So we counted a real one, and this note sets out exactly what we counted, what gets printed, and where it stops being reliable.

The rule that shaped it: the block prints even when the listing is silent

There is a design decision underneath all of this, and it came from a measurement that went badly for us.

We had assumed the heart of the €99 tier would be working out which business is behind an anonymous advertisement. Then we measured it properly, on twenty listings chosen so the mix of professional and private sellers matched the real market: it worked three times out of twenty. Building a paid product on something that happens once in seven attempts would mean two buyers paying the same money and receiving documents of wildly different weight.

So the layer is built from what can be produced for any listing we can open. And it follows one rule without exception: every block prints, even when the advertisement gives no number. If the listing does not state the price, the block prints what the market charges and how many listings do state it. There are no blank lines — only lines about what this listing chose not to say.

That inversion turns out to be the useful part. A buyer staring at a silent advertisement cannot tell whether the silence is remarkable. The layer answers that specific question.

Every share is printed with the number behind it

Look at any figure in the report and you will find a count in brackets after it: (140 listings), (64 listings), (54 listings).

This is not decoration and it is not modesty. A percentage with no sample size attached is an opinion wearing a disguise. "Eighty per cent of buyers" could come from four people.

We can now show you what that costs, because it happened to us. Until 13 August 2026 the bars-and-restaurants row rested on 22 listings, read off the first page of the section, and its middle was €44,000. We went back and read six pages of bars and six of restaurants — 226 distinguishable offers instead of 22 listings — and two things came out of it. Bars and restaurants are not one market: the middle bar asks €60,000 and the middle restaurant €129,000. And the old joint figure of €44,000 was not slightly off, it was low by more than half. The number was correctly calculated the first time; the sample was simply one screen, and one screen is not a sample. That is why the count in brackets is there, and why we would rather hand you that judgement than make it for you silently.

The same goes for the middle. We never print an average, because one tobacconist at €425,000 drags an average away from anything you would recognise. Instead we print two things: the middle 50 % — line the listings up cheapest to dearest, cut off the cheapest quarter and the dearest quarter, and print what is left — and the value sitting exactly in the middle of the line.

The seven measures, and the numbers behind them

1 · How long it has really been for sale

Not the date the portal displays. The date it first appeared.

Ten years. Somebody has been trying to sell that business since before the pandemic.

Time on the market is not proof of anything on its own — a good business can sit unsold because the asking price is wrong, or the owner is not in a hurry. But it is the single most reliable thing you can learn about a listing without knowing whose business it is, and it is the thing the portal's own display is designed to hide.

2 · Whether "refreshed today" is the real date

96 % of listings had been refreshed that day or the day before. Almost everything on the screen says it is new. A listing that has been quietly unsold for fourteen months appears to you as "posted 3 hours ago", and that is not an accident of the software — it is the point of the refresh button.

Where the two dates differ, the report prints both. Where they genuinely match, it now says so in words: published and refreshed the same day — here the visible date matches the real one. That last sentence is there because of a mistake, described further down.

3 · The asking price against its own sector

Those rows were re-counted on 12 and 13 August 2026, section by section, and they replaced a shorter set counted on 2 August. Three sectors deliberately have no row of their own. Pharmacies: only 13 of 157 pharmacy listings state a price at all, so a row would describe the few who talk rather than the market. Gyms: 20 listings with a price, too few. Hotels and guesthouses: the portal keeps them in one section, and asking prices there run from four figures to nine — the middle of that is not a market. Listings in those three trades are measured against "everything else", and the report says so on the line where it does it.

Two of the rows are a section of the portal rather than a trade, and we would rather say so than let the label pass. Dry cleaners and laundrettes share one section, and they are not the same business: the ones with a solvent machine ask around €75,000, the self-service laundrettes around €40,000, on nine and thirteen priced listings — too few for rows of their own. Academies are worse: one section holds language schools (19 offers, middle €35,000 on 16 prices), exam and tutoring centres (14 offers, middle €17,500 on 12 prices), and twenty more listings covering dance, music, beauty and hairdressing training, computing, one driving school and one nursery. Neither sub-trade reaches the 25 priced listings we require, so each gets one row with an honest label. The alternative was worse: without a row, an academy would be measured against every transfer in Spain, whose middle is €80,000 — off by three times instead of two.

The report says where this listing falls in its row — that is more than about 70 out of every 100 comparable listings ask — and never says whether that is good. Position in a line is a fact. "Expensive" is a verdict about a business we have not examined.

That sentence used to be worded "priced above roughly 70 % of them", and it was changed the day this note was written, because four independent readers all pointed out that it can be read as 70 % more expensive. That is not a shade of meaning — it is an error of several times over, in the direction of alarming you.

4 · Price per square metre

Across all listings that state both price and floor area, the middle half runs €333 to €1,223 per square metre, with €633 in the middle (243 listings). For bars and restaurants specifically it is tighter: €286 to €716, with €400 in the middle (21 listings).

5 · The price expressed in years of its own rent

This is the measure I would keep if I could keep only one.

Take the transfer price and divide it by a year of the rent on the premises. The middle half of the market pays between 3 and 8 years, with 5 in the middle (54 listings).

Why it is the tightest measure we found: rent is the price of the location. A busy corner in Madrid costs more per month than a side street in a small town, and that difference is already inside the rent figure. Dividing by it strips out most of what makes two businesses incomparable, and what remains is closer to a like-for-like number than any raw price can be.

It is also the measure that most often makes a buyer stop. A transfer priced at eleven years of rent is not illegal or dishonest — but it is well outside what the rest of the market is paying for the same kind of thing, and the seller should have an answer for it.

6 · Who is actually selling

Tracing the sellers behind 511 listings, held by 241 different accounts:

• 81 % of listings came from professional accounts, not owners.
• 62 % came from sellers running more than one listing.
• Ten sellers held 43 % of the entire section.
• The largest single seller was running 105 listings alone.

This changes how you should read everything else. When you write to a listing you are usually writing to an intermediary with a hundred other files open, not to a person who has stood behind that counter for eleven years. It also explains why the photographs so often show no shopfront: an intermediary has no reason to make their listing identifiable, and several reasons not to.

7 · What the listing does not say — next to how often others say it

Of 140 listings read line by line:

This is the least obvious block and, I think, the most useful, because it works against a natural instinct. Buyers get suspicious about the missing revenue figure — and the missing revenue figure is simply how this market behaves. Nine listings in ten stay silent about it, and that silence tells you nothing whatever about your seller.

Meanwhile the missing rent, which four sellers in ten do state, goes unremarked — and that one is genuinely out of the ordinary. The layer moves your suspicion from where it feels natural to where it is earned.

Where a field is missing and almost nobody states it, the report says so in the same breath and adds where the number actually comes from: the seller's own documents, at a tier that reads them.

Why the wording goes softer at the edges

When a listing lands in the bottom or the top tenth of its row, the report changes what it says. It prints the position and then adds, in fixed words:

"It sits at the edge of the range, where listings differ for reasons a listing does not show. Read this as a question to ask, not an answer."

The reason came out of measuring how far apart the prices actually spread. Near the middle, listings are broadly comparable. Out at the ends they are not: between the tenth listing in the line and the ninetieth, the gap runs as wide as seventeen times over. At that point the number is being driven by things no advertisement reveals — a lease with two months left on it, equipment worth more than the business, a licence that cannot be transferred to anyone.

The important part is that this softening is written into the text the engine prints, not left to whoever reads it. A rule that depends on someone remembering to be careful is not a rule.

Where the base stops — and this is printed in your report too

• Everything here comes from one website, in one country, collected on two or three days — the transfer section of a Spanish marketplace: how long listings stay up counted on 2 August 2026, the price rows re-counted on 12 and 13 August. A Portuguese listing is still measured against that Spanish base, and the report says so on the page rather than letting the figures pass as local.
• Seasonality is not measured. One snapshot cannot tell you whether August looks like February.
• A sector row is really a section of the portal. The sector comes from the portal's own category list, so the bakeries row contains a share of pastry shops and the hair-salon row a share of nail bars — where the section carried enough detail we dropped what did not belong, and where it did not we left the section as it stands and label the row as the portal's own section. Three trades have no row at all, for the reasons above.
• Duplicates: removed in the price rows, not in the rest. In the price rows the same offer republished on the next page, or by a second agency under the same wording, is counted once — that is what "distinguishable offers" means. In the exposure and disclosure figures, which still come from the 2 August count, duplicates were not removed, and with 81 % of listings coming from professionals that is not a theoretical worry.
• Abandoned listings are not separated from live ones. The advertisement that has been up for 3,918 days may not be a business that has been for sale for ten years; it may be one that was forgotten. That inflates the old end of the exposure figures by an amount we cannot state.
• Counted by us, not quoted from anyone. There is no industry survey behind these figures. That is a strength for honesty and a weakness for breadth, and both halves of that sentence are true.

The strongest objection: 344 listings on one day is not "the Spanish market"

It is not, and I want to put that argument at its full strength rather than in a version I find easy to answer.

One website is not every website. One day is not a year. Advertisements are not businesses — some are duplicates, some are relistings, some are abandoned. Printing sample sizes does not turn a narrow base into a representative one; it just makes the narrowness visible. Anyone using these figures to state what businesses in Spain are worth would be building on sand.

What I accept. The duplicate problem and the abandoned-listing problem are real, they are unmeasured, and I have added both to the list above rather than leaving them for a reader to work out. If these numbers were being sold as a valuation input, that would be disqualifying.

What I do not accept. They are not sold as that, and the frame of reference is narrower than the objection assumes. You are not choosing between all businesses in Spain. You are choosing between the listings in front of you on the portal you are actually shopping on, and for that specific comparison a full count of that portal's section is close to the right thing to compare against — better, not worse, for being a census of the place you are standing rather than a sample of the country. The question the layer answers is "how does this one sit among the others I could click on today", and it says so in those terms.

Why I have not simply widened it. The obvious fix is to sweep several portals every day for three months. We do not do that, and it is a rule rather than an omission: we read pages one at a time, like a visitor, and do not run bulk collection against portals whose terms forbid it. A wider base bought that way would be a base I could not defend, which is a worse problem than a narrow one I can describe.

So the honest position is this. These are the best figures I have, they are counted rather than guessed, their limits are printed alongside them, and where a decision would turn on their breadth rather than their existence, they should not be the thing you lean on.

Two things a live run caught, and what they cost

When we first ran the finished layer against a real listing, it produced two wrong outputs. Both are worth describing, because they are the kind of fault that never announces itself.

A bar was compared against the wrong row. The part of the engine that reads a listing labels its sector with one vocabulary; the part that picks the comparison row expected a different vocabulary. Neither crashed. The bar was simply measured against "everything else" instead of against bars and restaurants — and since those two rows have different middles, the conclusion about its price came out different. A silent mismatch between two lists of words changed the answer.

A listing published that same morning was reported as saying nothing about its freshness. It had nothing to say, because published and refreshed were the same day — but printing "the listing does not state it" reads as a gap in our work rather than as the actual answer. Now it says the true thing in words.

Both are fixed and both now have tests. I mention them because a page describing a measuring instrument should also say what happened the first time it was pointed at something real.

What this layer will never do

• It does not value the business. It gives a position in a line and the size of that line.
• It does not say "expensive" or "cheap". Those are conclusions about a business it has not examined.
• It does not verify a single claim in the listing. If the advertisement says the rent is €1,200, the arithmetic uses €1,200. Testing that figure needs the lease, which is a different tier.
• It cannot see a relisting. If a seller deletes an advertisement and posts it again, the age restarts at zero and we will report it as new, because to the portal it is new.

All figures in this note come from our own count of the transfer section of one Spanish marketplace on 2 August 2026: 344 listings measured for age and price, 243 for price per square metre, 140 read line by line for the disclosure shares, 54 with both a price and a rent, 511 traced to 241 seller accounts, and 20 taken in market proportions for the identification test. Sample sizes are printed with every share, in this note and in every report. Everything comes from one website, in one country, on a single day; seasonality is not measured; sector rows are narrow because two-thirds of listings sit in the portal's general "other" group; duplicate advertisements were not removed; and abandoned listings were not separated from live ones. General information and commercial opinion, not legal, tax or financial advice.

Provenance Diligence — independent, buyer-side screening of small-business purchases in Spain & Portugal. Published notes and Deal Screen / Reality Check memos are personally reviewed and signed by the founder (who I am →). I take no commission and never contact the seller.