The method

Why VERA exists, and how it works.

01The story

Every number we were given was somebody’s position. So we built our own.

VERA was not born of a business plan. It came from looking at villas on Cap d’Antibes and finding that nobody in the room could say what one was actually worth – only what they hoped it would fetch.

IThe wall

Asking prices came with confidence and no working. Comparables arrived hand-picked, three at a time, always flattering. Ask what the house next door actually sold for and the conversation moved on.

IIThe public secret

The strange part is that the answer is public. The French state records every completed sale and publishes it. But it arrives as raw transaction files and cadastral geometry – no addresses you would recognise, no floor areas you can trust, no way to tell a villa from a garage. Open data that almost nobody can read.

IIIThe engine

So we built the thing we wanted to be handed: an engine that reads the register for one peninsula, corrects for size and location, and states a value with the evidence attached. We finished it, used it, and realised that anyone buying here is facing the same wall. That became VERA™, the property engine of VOR Media Technologies. Cap d’Antibes is its first market.

The transaction register as published: forty columns, one line per lot, addresses as road codes — The register, as published. Forty columns, one line per lot, addresses as road codes, the same sale repeated for every cellar and parking space attached to it. Nothing is hidden – it is simply published in a form that makes it useless to the person it would help most. What stands behind every VERA number is this file, and you can check every line of it.

Five things VERA will not doPromise a discountHide the workingsValue a market it has not built the evidence forKeep the method to itselfTake an interest in the transaction

We do not control the vendor, and a valuation that flatters the buyer is worth as little as one that flatters the seller. Every recorded sale used is listed in the Valuation Brief, including the ones that argue against our own conclusion. If the transaction record for your area is too thin, we say so rather than extrapolate. The whole valuation is reproducible from three public files and one script. We are not an agent, we earn nothing from the sale, and we are not paid more if you buy.

07Asking price vs evidence

The asking price is the seller’s case. VERA builds yours.

An asking price is a commercial position: what the seller hopes to achieve, supported by the case made to the market. VERA tests that position against completed transactions and turns the evidence into an opening offer, a defensible range and a walk-away point. Both have a place in the negotiation. Only one is prepared for you.

01The asking price opens the negotiation

It reflects the seller’s expectations, the agent’s judgement and the strategy for bringing the property to market. It tells you where the seller wants to begin – not where the evidence says you should end.

02A few comparables are not the market

Nearby properties can make a persuasive case. VERA tests that case against the full qualifying evidence set, including the transactions that weaken it as well as those that support it.

03Evidence turns doubt into a position

“It feels expensive” is easy to dismiss. A property-specific valuation tested against 221 qualifying comparable sales gives you an opening offer, a defensible range and a clear point at which to walk away.

The evidence

One price asked. 194 prices paid.

An asking price is one commercial position. The French transaction register records what buyers actually paid. VERA identifies every qualifying house sale on the peninsula, adjusts each one to the villa’s floor area and plot, and maps what the surrounding evidence supports at that location.

Qualifying completed house salesThe villa being analysedThe local rate supported by surrounding transactions · €11,000–€16,000/m²

Local evidence rate€14,000/m²

Asking rate€22,862/m²

Premium to local evidence+63%

At this address, the local evidence rate is approximately €14,000/m². The asking price implies €22,862/m² – about +63% higher. The map is not the final valuation. It shows why the asking price should be tested before you name yours.

Each qualifying sale is adjusted to the subject villa for floor area and plot size, then weighted by distance. The colour surface shows the local median rate only where sufficient evidence exists. Sources: Demandes de Valeurs Foncières, DGFiP, and the Etalab open cadastre.

See the sales behind the numberA dozen of them, so you can see they are real

Qualifying sales on the peninsula, at the price entered by the notaire – a dozen spread evenly across the range, with the top and bottom tenth left out. As recorded is the price divided by the floor area in the register. Restated is that rate adjusted onto this villa for floor area and plot, which is what the valuation works from; it moves some rates up and others down.

Metres away sets how heavily a sale counts, not how dear it should be. The nearer a sale, the more weight it carries in the local rate. It is not a price signal in itself: across all 194 sales the correlation between distance from this villa and rate per m² is close to nil, because the expensive ground is the eastern shoreline rather than a ring around any one address. A sale a kilometre north along that shore can be dearer than one next door but inland.

The tails are left out because they are not price levels. The register holds floor area as declared for tax, so the dearest sale in the full set reads €128,500/m² only because a €25,700,000 estate carries 200 m² of declared area, and the cheapest carry the same fault in reverse. They stay in the valuation, where the adjustment handles them. The full schedule in the Valuation Brief runs to every one of the 194.

The twelve nearest qualifying sales, street by street with the rate as recorded, are listed in the Brief itself – individual deeds are not published on this site. Source: Demandes de Valeurs Foncières. Distances are straight-line from the villa and set each sale’s weight in the kernel. Restatement adjusts for floor area and plot only – never for condition, which the register does not record.

See how 39,035 rows became 194The 194 is what survived

It sounds thin until you see what was read to find it. The engine starts with every property transaction recorded anywhere in the commune of Antibes over 5 years, and discards, in public, everything that cannot honestly be compared to a villa on this peninsula.

Most of what a naive reading would count is not a house at all. One sale appears in the file once for every lot attached to it – the flat, the cellar, the parking space – so the largest single conveyance in the period occupies 339 separate lines. Add those lines together and you get €39.9 billion of apparent turnover in a commune that actually transacted €5.90 billion. Anyone who sums the column without collapsing it first is out by a factor of 6.8.

39,035rows downloaded from the register, Antibes 06004, 2021–2025
36,432after removing exact duplicate lines−2,603
15,313distinct conveyances, once every lot is collapsed onto its deed−21,119
14,282sales rather than gifts, divisions or expropriations−1,031
2,560within two kilometres of the villa−11,722
310involving a house at all, rather than a flat or a plot−2,250
277a single house – not 30 deeds bundling several, nor 3 pairing a house with a flat−33
273priced like an arm’s-length sale−4
224of 100 m² or more−49
221after 1 duplicate deed and 2 with no plot area recorded−3
194on the peninsula itself – the evidence this valuation is built from. The other 27 lie just beyond it, inside the two-kilometre ring.

Every exclusion is listed by deed reference in the Valuation Brief, including the 49 houses dropped for size or price, so the sift can be disagreed with rather than taken on trust. The peninsula is also mapped against the full cadastre – 16,624 parcels and 32,361 buildings across the commune – and the distance weighting is not chosen by hand but picked from 54 candidate settings by leave-one-out testing. The valuation reads five years; VERA holds twelve. 3,413 recorded conveyances across 845 parcels back to 2014 drive the parcel and street history in the Pre-offer Brief. The valuation itself uses the most recent five because the older years add dispersion rather than signal – tested against a three-year window, five costs 2.7% of cross-validated accuracy and buys three times the sales on this villa’s own street. The wider interval is the stated consequence, not a rounding.

Fall for the villa. Negotiate from the evidence.

08How VERA works

Six independent readings. One defensible answer.

VERA – Verified Evidence, Research & Analytics – collects the public evidence, values the property six separate ways, and states what it does not know. On the Cap d’Antibes, 194 recorded house sales 2021–2025 qualify for the valuation model.

COLLECTPublic sources, read automatically

The transaction register, the cadastre, the address base and the census, assembled parcel by parcel and matched to a location rather than a postcode.

MODELSix readings, not one opinion

An adjusted comparable grid, a matched-band median, a peninsula median, the same-street evidence, a hedonic regression and a third-party automated estimate – each run separately, each weighted on how much it deserves to be believed.

QUANTIFYA range, with the odds attached

One million simulations propagate the uncertainty in area, condition and method error into a value range, so the Valuation Brief gives you a probability rather than a single confident number.

Every figure in a VERA Valuation Brief is traceable to a public source or to a stated assumption. Where the evidence is thin, the Valuation Brief says so.

The simulation

Run a million times. It never reached the asking price.

A single figure hides how much is unknown. So the engine runs the whole valuation a million times, varying everything it is not certain about, and gives you the odds instead of a number it cannot defend.

Even on our own worst error measure, the asking price is only 5% likely.

1,000,000simulated valuationsEach draws a habitable area, a condition allowance, a view credit and a correlated error for every method

0reached the asking priceNot one trial in a million reached it – the highest outcome the model can produce at all is €4,426,000, €74,000 short of the ask

50%landed in the stated rangeWhich is what the range means – half the outcomes, not a promise

See the six methods, their values and their weightsSix methods, six answers

Method	Value	Weight
Adjusted comparable grid	€3,243,000	27%
Hedonic model, Duan-corrected	€3,433,000	18%
Matched-band median	€3,360,000	16%
Peninsula median, plot ≥800 m²	€3,439,000	16%
Nearest evidence, within 500 m	€3,175,000	16%
Automated valuation model	€3,117,000	7%

They disagree, by €322,000 from lowest to highest, and that disagreement is itself information – it is where the range comes from. The important part is that the highest of the six still lands €1,061,000 below the asking price.

Weighted together the six read €3,308,000. The published central value of €3,326,000 is the median of the simulation, which also varies floor area and condition – the two differ by less than 0.6%, and we print both rather than only the one that flatters.

What moves the number

Five things carry the uncertainty. The Valuation Brief ranks them.

Every valuation rests on assumptions. Most reports bury them. This one measures how much each is worth, so you know which arguments are worth having and which are noise.

Comparable evidence · 95% intervalSampling error across six methods€574,000 of range

Condition and works−12% to 0% against a mixed-condition median€425,000 of range

Habitable area basis · 230 vs 253.7 m²Unreconciled 23.7 m²€317,000 of range

Planning and heritage zone−3% to +2%€167,000 of range

Sea view credit0% to +5% · partial view only€163,000 of range

€3.02mCentral value€3.64m

How each band was measured, and why they do not add upThe one you can do something about is the floor area

The largest band is the evidence itself – €574,000 of sampling error across the six methods, which no amount of care removes. It is the honest cost of a market that trades fifty houses a year.

But the third band is a choice, not a fact. The advertisement said “about 230 m².” The surveyor’s certificate measures 196.83 m² under loi Carrez, plus 56.85 m² the law requires be excluded. Twenty-three point seven square metres are unreconciled between the two, and on this house that is worth €317,000.

Which is the whole argument for reading the certificate rather than the listing. A number nobody checked is not a small problem here – it is the third largest source of doubt in the valuation, and the only one a buyer can resolve before signing.

Each band is the value the model returns when that one assumption is moved across its plausible range and everything else is held at its central setting. The bands are not additive: the simulation varies all five together, which is why the published range is narrower than stacking these end to end would suggest.

11The alternatives

Three sources of a number. Only one is built for the buyer.

Each was produced for a different purpose, and only one of them was produced for you. VERA comes at three depths – a free scale on this page, a Pre-offer Brief for the villas on your shortlist, and a Valuation Brief for the one you decide to bid on.

How an agent’s valuation, an online estimate and VERA at its three depths differ
	The agent	An online estimate	VERA, free	VERA Pre-offer Brief	VERA Valuation Brief
The question it answers	What the seller hopes to achieve	What houses like this tend to fetch	Is the asking rate high for this peninsula?	Is this one worth a closer look?	What is it worth, and what should I offer?
Based on	Asking prices, and the agent’s own judgement	A national model, applied street by street	The same 194 qualifying peninsula sales	194 qualifying peninsula sales, and the five nearest deeds	221 qualifying comparable sales – 194 of them on this peninsula
Produced for	The seller’s instruction	A mass-market audience	Anyone, before they speak to a soul	The buyer, while the shortlist is still long	The buyer, and nobody else in the transaction
Gives you a price to offer	No – it is the price to beat	No	No. It knows nothing about the house itself	No, deliberately. It does not value the house	Five: opening, floor, central, ceiling, walk-away
Shows its working	No	No	Every recorded rate, drawn on one scale	Every deed it cites, mapped, with the walk in metres	Every sale listed, every method weighted
States its own error	No	No	No – it is a position, not a valuation	The interval on the local rate, and how many sales carry it	Yes – 25% out of sample, against 30% for the rule of thumb most buyers use
Quantifies uncertainty	No	A confidence badge	No	A percentile, not a point	1,000,000 simulations, as a range
Use in a negotiation	It is the other side’s position	Too generic to argue from	None. It is the question, not the answer	Tells you whether to open one at all	Written to be handed across the table
Turnaround	On request	Instant	Instant, on this page	24 hours from the address	24 hours from confirmation
Cost	Free	Free	Free – try it above	€2,000 – credited against a valuation	from €4,500

How an agent’s valuation, an online estimate and VERA at its three depths differ
	VERA, free	VERA Pre-offer Brief	VERA Valuation Brief	The agent	An online estimate
The question it answers	Is the asking rate high for this peninsula?	Is this one worth a closer look?	What is it worth, and what should I offer?	What the seller hopes to achieve	What houses like this tend to fetch
Based on	The same 194 qualifying peninsula sales	194 qualifying peninsula sales, and the five nearest deeds	221 qualifying comparable sales – 194 of them on this peninsula	Asking prices, and the agent’s own judgement	A national model, applied street by street
Produced for	Anyone, before they speak to a soul	The buyer, while the shortlist is still long	The buyer, and nobody else in the transaction	The seller’s instruction	A mass-market audience
Gives you a price to offer	No. It knows nothing about the house itself	No, deliberately. It does not value the house	Five: opening, floor, central, ceiling, walk-away	No – it is the price to beat	No
Shows its working	Every recorded rate, drawn on one scale	Every deed it cites, mapped, with the walk in metres	Every sale listed, every method weighted	No	No
States its own error	No – it is a position, not a valuation	The interval on the local rate, and how many sales carry it	Yes – 25% out of sample, against 30% for the rule of thumb most buyers use	No	No
Quantifies uncertainty	No	A percentile, not a point	1,000,000 simulations, as a range	No	A confidence badge
Use in a negotiation	None. It is the question, not the answer	Tells you whether to open one at all	Written to be handed across the table	It is the other side’s position	Too generic to argue from
Turnaround	Instant, on this page	24 hours from the address	24 hours from confirmation	On request	Instant
Cost	Free – try it above	€2,000 – credited against a valuation	from €4,500	Free	Free

Register last read 28 September 2026 · 194 qualifying peninsula sales · next refresh when DGFiP publishes. The agent’s role is to market and sell the property. VERA’s role is to give the buyer an independent view of value. A good agent is worth knowing and will tell you things no register holds – but their figure was produced for a different purpose than yours, and it should not be the only one you have.

12Beyond the price

The price is the smallest decision you make.

Two things decide whether a house on Cap d’Antibes was a good idea, and neither of them is visible at a viewing: what it costs to keep, and what the certificates already say about it.

A large villa with pools, courts and lawns on Cap d’Antibes — Everything in this photograph has a running cost. The pools, the courts, the lawns that should not survive a Mediterranean August, the staff who keep them – none of it appears in the asking price, and all of it appears every year.

What it costs to keep

Modelled low, central and high – wealth tax, local taxes, insurance, utilities, pool and grounds, a caretaker, a maintenance reserve and compliance.

Tax€32,500

Staff and grounds€15,000

Maintenance€15,000

Insurance and utilities€6,500

Running€5,500

Ten years, at the central case€744,000

22% of the property’s value, before a single improvement. Annual range €50,000 to €119,000.

Read the diagnostic file, certificate by certificateWhat the certificates already say

The vendor’s diagnostic file is disclosable and usually skimmed. Read properly, it is a list of things that are about to become yours.

Three of the six certificates have lapsed – the lead report, the termite report and the risk statement, which must be under six months old at the promesse.
Energy class E at 280 kWh/m²/yr. Class E carries a letting prohibition from 1 January 2034 – irrelevant if you live in it, decisive if you let it.
Four electrical anomalies, including a pool circuit with no 30 mA protection, and nine control points the inspector could not see.
Lead present in four of 167 tested units – four painted grilles, ordinary wear, not a crisis, but it is why the report has lapsed.
Prescribed fire-risk works on the parcel, with completion left unanswered on the certificate. A geotechnical study is compulsory here before any construction, under loi ELAN article 68.

None of this makes the house a bad buy. It is a house of the 1920s and it reads like one. The point is that all of it is knowable before you sign, all of it is priceable, and every item above is either a condition to attach to the offer or a number to take off it.