First analysis — second one free. All September!

See pricesPrices
News

The market price estimate just got more accurate

The market price estimate just got more accurate

The market price is one of the most important numbers in an m²cena report. Much of the rest of the analysis is built around it — where the flat stands in the market, how it compares with other properties, and in the end the decision whether it is worth a closer look.

That is why we check the market price calculation particularly carefully. This time the check ended with us changing not individual details, but the very way the market price is calculated.

What changed

The market price is now produced by a model, not by the average of similar deals.

The estimate used to rest on the average price of similar deals nearby, and every flat was put through much the same recipe.

The market price is now produced by a model trained on real property transactions. It works out which factors help most in a given situation to establish a flat's value. If the building itself has enough sales history, those deals carry more weight. If it does not, the model leans more on deals in the immediate area.

The building is read in far more detail.

The calculation takes in the build year, the exterior wall material, the floor the flat is on, the building's energy rating, the renovations done and other factors. What matters is not only each feature on its own, but their combination.

That means two buildings on the same street are no longer valued as practically identical objects just because they stand next to each other.

The effect of a flat's condition on the price has been recalculated.

We still take the flat's condition into account, read from the listing photos. But we have revisited how much it moves the calculated market price.

In newer buildings, good condition used to raise the estimate too much. Yet in a new building a good interior is often expected already and is partly reflected in the value of the building itself. You should not pay twice for one and the same advantage.

Now the model awards extra value for condition mainly where a flat genuinely stands out against comparable flats.

Area for the calculation comes from the cadastre.

The area given in a listing does not always match the officially registered one. A listing may count a loggia into the area, for instance, while the cadastre does not include it in the flat's area.

In such cases we use the area registered in the cadastre for the market price calculation. The report shows both numbers — the one from the listing and the one available in the register — so the difference is visible at once.

What it means for you

Overall, the new estimate is closer to actual transaction prices. The more important improvement, though, is that the calculation became more accurate exactly in the cases where it used to miss most often — in newer buildings and in flats in very good interior condition.

So for some listings the market price estimate will have changed. In a few cases the report's verdict on the asking price may change with it.

A market price estimate is not a promise that the property will sell for exactly that sum. It is our best data-based estimate of the price this particular flat could sell for right now.

If you open an analysis you created earlier and see a different market price estimate than you remember, that is expected. The report always uses the latest version of the calculation rather than storing the estimate as it stood on the day the analysis was made.

How we checked the new model

We check every change to the calculation by one principle: we take flats whose actual sale price we know, and check what market price the model would have calculated for them before the deal, without seeing its final price.

We then compare the model's estimate with the actual transaction price and with the results of the previous version.

If the new version does not show better results than the previous one, it does not reach users' reports.

If an estimate still looks strange to you, write to us. Examples like that help us find the situations where the model can still be improved — and often enough, that is where the next change starts.

Searching listings…

Preparing your report