MyHomePrice

Methodology

Making property expertise available to everyone — built on data, not on sales talk.

Analysing a property used to mean an appointment and several hundred euros. We build it on official French public data and open models, so anyone can decide with the same rigour as a professional. Here is how every figure is computed — and why you can check it.

2014–2025

of real DVF sales

All of France

official DGFiP data

IRL & ECB

models anchored to official indices

What makes our figures reliable

Real sales, not listings

Every figure rests on transactions actually recorded by the tax administration (DVF, DGFiP), 2014–2025 — the same data notaries and public authorities use, not asking prices that never close.

Repeat-sales index (Case-Shiller type)

We track the resale of the same property to isolate the market's real trend at constant quality — the reference method of the major house-price indices, applied to France.

A forecast that reverts to the mean

An area's past growth does not go on forever: our model converges the local trend towards the French long-term norm, with scenarios anchored to real interest-rate paths (ECB).

Rents indexed by law

Rents (official Carte des loyers, 2022–2025) are revalued with the IRL — the legal index that caps increases at roughly inflation. No inflated yields: what the law actually allows.

Published, verifiable models

No black box. Our methods and their error margins are measured and published below — you can judge the reliability rather than simply trust us.

Our conviction

Online property valuation is broken: most “free estimates” are lead-capture forms sold on to agencies that call you for weeks. MyHomePrice does the opposite: a 100% online service, no agency, no intermediary, based exclusively on sales actually signed at the notary. Nobody will call you. Your report is ready in one minute.

The data: real sales, not listings

Everything rests on the DVF database (Demandes de Valeurs Foncières): every transaction recorded by the tax administration since 2014, published as open data by the French State (Etalab). 9.2 million house and flat sales analysed, nationwide, updated automatically at each official release (April and October). No asking price enters our calculations: signed prices only.

Current value estimate

Median price per m² of real sales in your postcode over 3 years, volume-weighted; weighted floor area (basement, garage, terrace); about twenty criteria calibrated with an architect's expertise — condition, energy rating, heating, party walls, kitchen and bathroom… plus judgements generic grids miss: in a Haussmann building the 3rd to 5th floors sell best, not the top one. The result is an honest range (±8% locally, ±12% when falling back to the département), with the sample shown and every coefficient detailed in the PDF report.

On top comes an “AI second opinion” (beta): a machine-learning model trained on 4 million comparable sales (2021-2025) values the same property from floor area, rooms, land and the local price level only — without seeing the condition or energy rating you declared. An independent cross-check with a published error: 18% median error on the test year, against 20% for the plain local median. When the two figures converge, confidence rises; when they diverge, it signals that your property's condition weighs heavily. Try the valuation →

Price forecast: statistics and artificial intelligence

Two engines, comparable in one click. The statistical model (our reference) starts from your area's real trend and converges it towards the French market's long-term pace (mean reversion) — with scenarios anchored to real rate paths: mortgage rates, inflation and the 10-year OAT published by the ECB, smoothed and recalibrated monthly.

The AI model (beta) is a gradient-boosting algorithm trained on 57,000 zone-year observations from 12 years of DVF. It forecasts the 1-year and 5-year horizons directly — mean absolute error of 6.5% at 1 year and 2.8% per year at 5 years on the test year, better than naive extrapolation. Beyond 5 years, where no data can decide, the path rejoins the market's structural pace: we publish our limits as well as our results. Try the forecast →

The MyHomePrice Index (repeat sales)

The median price distorts when the mix of sales changes. To measure the true trend we built a repeat-sales index — the method of the famous US Case-Shiller index: among 9.2 million transactions, 1.25 million cases where the same property sold twice (matched by cadastral parcel), Bailey-Muth-Nourse regression, published for France and 96 départements (base 100 = 2014). The index captures the 2021-2024 Paris correction and the Bordeaux pull-back alike — movements that averages smooth away. See it on every département page. Paris (75) →

AI analysis of the diagnostic file

Upload the diagnostic file (DDT) or the sale agreement with its annexes for the property you are targeting: the AI extracts the ranked critical points, repair costs, your negotiation levers and the questions to ask the seller — with the file's page numbers referenced so you can check everything in the original. In tests on real files, the analysis caught genuine clerical errors (a cadastral area shifted by one column between the deed and an annex).

What we do not do

  • ❌ No canvassing: your valuation triggers no agency call.
  • ❌ No data resale: your information never leaves MyHomePrice.
  • ❌ No asking prices: signed transactions only.
  • ❌ No yield promises: bounded scenarios, never a guarantee.

In short

MyHomePrice is one of the first consumer services in France to combine AI property forecasting, a département-level repeat-sales index and scenarios anchored to real rates — applied to public notarial data, with a result in one minute and no intermediary. Tools once reserved for professionals, now open to everyone: free for the figures, €9.99 when you want the AI's opinion on a specific property.

MyHomePrice estimates and forecasts are provided for information only, from public data (DVF/Etalab, ECB). They are neither a property appraisal nor investment advice.