Reconstruction value
The cost of rebuilding the structure under the statutory valuation rules, the base for the technical value and for insurance.
Bytero Engine™
An automated valuation model (AVM) for flats and houses that shows the comparables behind each value.
Enter the address, the floor area, the year of construction, the floor and the number of storeys. The AVM returns the market value, the range around it and the comparable properties it was priced against. For a house the living area is required and the condition is asked; a house value is less precise than a flat value, and the page says so beside the figure.
Get accessPrices, rents and values
What comes back for an address
The market values the AVM reads from current sale and rental prices, and the statutory values an appraiser or an insurer needs, each per address and all through the API.
Market value€
The level at which comparable flats are offered today: a regression averaged with the nearest comparable properties, each adjusted for size, floor and year. Where no comparables exist the regression stands alone and the answer says so.
Asking price€
What comparable properties are offered for right now in the same location and segment.
Rent€ / mo.
Achievable monthly rent from a second model trained on rental prices.
Yield% p. a.
Gross annual rental yield: predicted rent times 12, divided by the predicted price.
Price per m²€ / m²
Price per square metre of usable area, comparable across layouts and cities.
Price trenddaily data
Daily price snapshots of each property on the market show drops, increases and how long it has been offered.
Statutory values
Technical value
The reconstruction value reduced by wear for age and condition, computed the way an appraiser does.
Official value
The technical value adjusted by the location coefficient, as an expert report states it.
Insured sum
The base for an insurance contract, derived from the reconstruction value. It guards against under- and over-insurance.
Method
How the AVM computes a value
01
One row per flat
The inputs are the flats and houses offered for sale and to let, read daily. The same flat offered through several channels, or re-offered at a new price, becomes one row, matched by its photos, position and geometry. Prices are asking prices: the land register does not publish what was paid, so the discount between asking and agreed price is estimated separately and shown as its own figure.
02
The year from the register
The year of construction is the input that moves a value most, and it is often missing from the market data. Where it is missing the engine reads it from the building’s register history: the first titles of the block, then the wave of first purchases, and only then older evidence. Each year carries the source it came from.
03
Location as nested cells
A flat’s position is read as nested cells of 250 m, 1 km, 4 km and 16 km. A cell with few comparables leans on its parent, so a street with three comparables borrows from its district, and a district with thousands stands on its own.
04
A regression and a comparables search
Two methods price each flat. A regression reads the inputs together with the building’s attributes. A comparables search finds the nearest similar properties, the same building first, adjusts each for size, floor and year, and leans on the regression where comparables are few. The served value is the plain average of the two.
05
The test on flats it never saw
Each version is scored on held-out flats whose year is stated, valued from location, area, year, floor and storeys alone. The errors are published per age band, so a buyer can see how the engine does on the stock they hold.
06
A range around the value
Each value carries a range, measured per segment on held-out asking prices rather than picked by hand. A wide range is the engine saying it is less sure, for example a flat whose comparables disagree with each other.
07
Houses
Houses have their own pool and their own model. The living area is required, the condition is asked, and the plot area from the register is a hint only. A house value is less precise than a flat value, so it is shown with its own, wider range and says so.
With the value
What comes back with the value
The figure is never alone. Each answer carries its inputs, its comparables and its version.
The inputs and their sources
Each input shows the value used and where it came from: the caller, the register or an estimate. A wrong input is visible before the figure is trusted.
The comparables
The properties the flat was priced against, with the distance or the same building, whether each is still on the market, its size, year, price and adjusted price. Each one links to its record.
Asking level and agreed level
The value is the level at which comparable flats are offered. The expected discount between the asking and the agreed price is a separate figure, so the two are never mixed into one number.
The model version and the date
Each value names the model version and the date of the market data it was computed from, so two valuations of one flat can be compared later.