Listings from three Slovak portals in one place, ranked by where you travel and what matters to you, with prices per m² and a price estimate.
MapaTrhu, 01
A search shaped around you.
The guide starts with the home you want. Set a budget and location before browsing listings, then add commuting preferences and what matters to you.
MapaTrhu, 02
See the home and its surroundings.
Listings are ranked by how well they fit you. Each one shows what fits, what does not and what is unknown. The map and list stay together, and the app distinguishes exact and approximate locations.
MapaTrhu, 03
Put the asking price in context.
Compare asking prices per m² by district on a map and over time. A price estimate for a flat or house shows a range and similar listings nearby. Listing history does not forecast future prices.
MapaTrhu, 04
Take your search with you.
The map and guide have their own mobile interface. Keep listings and locations within reach when you are away from your desk.
- Listings from three Slovak portals, with copies of one flat merged into one listing
- A guide covering location or a drawn area, commute, priorities and deal breakers
- Listings ranked by your places, budget and priorities, each with a reason
- Minutes from each flat to a stop, a shop and the places you travel to
- A listing map linked to the results, also on a phone
- Saved searches with an e-mail when new matches appear
- A price map and flat and land prices per m² by district
- A price estimate for a flat or house, with a range and similar listings nearby
- Mortgage calculator, property tax and sale tax calculators
- Agency tools with price analyses shared by link, a weekly territory watch and own listings
Why the product keeps changing
The product grew from a shared map into a data platform, then a simpler everyday interface. Each stage addressed a different obstacle in property search. Screens shown here belong to V3.
V1
Bring listings together
The starting point was fragmented property search. A shared map made it possible to view listings alongside their location.
V2
Build the data foundation
A list of adverts cannot explain a price on its own. The next stage developed the data work needed for more useful comparisons.
V3
Less friction at the surface
The current version starts with a short guide and ranks listings by a person's places, budget and priorities. Behind the interface it merges adverts, calculates commutes and uses trained models whose limits stay visible. Tools for agencies came with it.
Numbers you can put in context
A product counter and a model evaluation answer different questions. These figures include their method and date. Live product totals will change.
- €/m² median flat asking price
2,869
Median asking price per m² for flats shown on the MapaTrhu price page.
A dated price-page snapshot of asking prices, not completed sales.
Method and source
A dated price-page snapshot of asking prices, not completed sales. A national median does not value an individual flat.
Source, MapaTrhu property prices
- flats and houses
37,770
Homes shown by the MapaTrhu homepage counter.
A dated counter snapshot, separate from sales volumes and the full advert archive.
Method and source
A dated counter snapshot, separate from sales volumes and the full advert archive.
Source, MapaTrhu public homepage
- source portals
3
Number of source portals reported by the product.
A public product snapshot in which multiple adverts may describe the same property.
Method and source
A public product snapshot in which multiple adverts may describe the same property.
Source, MapaTrhu public homepage
- % median error
8.7
Offline evaluation of asking-rent estimates from rent/v1.
An offline test on 703 later listings measures asking-rent estimate error, not agreed rents or future prices.
Method and source
An offline test on 703 later listings measures asking-rent estimate error, not agreed rents or future prices. Training used 7,036 earlier listings. The comparison rule scored 10.7%.
- training texts
2,863
SlovakBERT training corpus for extracting advert details.
The count of texts labelled by Claude Opus is not an accuracy score or expert validation.
Method and source
The count of texts labelled by Claude Opus is not an accuracy score or expert validation. Evaluation used 300 held-out texts.
Built together by Rise.sk and Wittlinger.dev
MapaTrhu is a joint product with Samuel Wittlinger. Its map interface and data workflows inform the custom software we build for clients.
The technology behind the product
Next.js and MapLibre power the interface. Python and FastAPI handle data workflows, with PostgreSQL and PostGIS for spatial data. LightGBM and SlovakBERT support estimates and advert extraction.
- Next.js
- React
- TypeScript
- MapLibre GL
- Python
- FastAPI
- PostgreSQL
- PostGIS
- LightGBM
- SlovakBERT
