About Prona
Prona gathers 79,148 live listings from MerrJep, DuaShpi.al, FutureHome, Njoftime.com and 6 other portals, processes them with AI so price, rooms, floor area and neighbourhood are right, and ranks them by price per m² against the median for their area. The result: filters that return what you asked for, and the genuine bargain first.
Why it exists
Anyone looking for a home in Albania has to open five or six sites, each with its own filters, and compare by hand listings that are often the same property at different prices. Even within one site the filters disappoint: a "€300" apartment for sale because the owner typed the price per m², "Don Bosko" spelled five ways, a rental filed under sales. Prona brings them into one list, with the same filters, corrected data and one unit of comparison: euro per square metre.
How the AI processes the listings
Every new listing goes through a language model that reads the original title and description, as the owner or agency wrote them, and compares them with the fields the source site recorded. Where the text clearly shows a field is wrong or missing, the model corrects it; where it is unsure, it leaves it alone. The fields it checks:
- Price and currency: per-m² prices become totals (€2,100 on 120 m² is €252,000), old lek is divided by 10, "42 thousand" recorded as 42 is fixed, placeholders like €1 or €123 are dropped, nightly rents are not indexed.
- Sale or rent, when the ad was filed under the wrong category.
- Rooms (2+1 is two bedrooms and a living room), bathrooms, floor area, floor, year built.
- City and neighbourhood under one standard name, so "tek Don Boskoja", "Rruga Don Bosko" and "Don Bosco" are one filter; Kamza, Vora and Fushe-Kruja as their own towns, not Tirana zones.
- Owner or agency, furnishing, lift, parking, building status.
- The most specific place the ad names (a complex, a street, a landmark), so the listing can be placed on the map.
- "Wanted" posts and ads that offer no property are removed from the results.
Every correction is stored field by field with the value before and after, and the price-per-m² statistics are computed only over checked listings, so an ad with a wrong price cannot skew the area median. Right now 76,551 live listings have been checked and 59,252 of them had at least one field corrected.
How the deal score works
For every listing the price per m² is computed and compared with the median of its segment: same city, same property type, same sale or rent category. A listing at 0.8 costs 20% less than the median for its area; one at 1.2 costs 20% more. When the source gives no floor area, the comparison falls back to total price for the same layout.
The default order starts with the lowest ratio, lightly penalised for the age of the listing, because a two-month-old bargain has most likely sold. Nobody can pay to be ranked higher.
What it is not
- Not an agency: we do not broker, take no commission and hold no listings of our own.
- Not a listing site: every ad belongs to the site where the owner or agency posted it and opens there.
- Not a valuer: the AI corrects a price only when the ad’s own text proves the recorded one wrong. Prices stay asking prices, not appraisals and not recorded sale prices.
What it covers
Sales and rentals of apartments, villas, houses, land, shops, offices and business premises in 40 cities, from 10 sources. The pages by city, area, type and layout need no account, and the full search adds filters for price, floor area, floor, lift, parking, furnishing and distance from a point on the map.
See also prices by city and the list of sources.
