A property listings site built from Facebook group posts
Property listings scattered across Facebook groups, turned into one searchable listings site.
The problem
In Addis Ababa, many rentals and sales never appear on a property portal. Agents and owners post them in Facebook groups instead: a few photos, a price, a phone number and a paragraph of text. Finding a flat means scrolling through dozens of groups, and the same flat reappears every few days, posted by different agents.
The client wanted a single site where those listings could be searched and filtered like on any property portal.
What we built
A pipeline that runs every two hours:
- Collect. New posts from 40 groups, with text, photos and post date.
- Filter. Most posts aren't listings. Questions, "looking for a two-bedroom" requests and promotions are separated out.
- Extract. Price, currency, sale or rent, bedrooms, neighbourhood and contact details, pulled from free text written in English and Amharic, often both in one post. The same price turns up as "25k", "25,000" or "25,000 a month, negotiable".
- De-duplicate. A flat reposted by three agents with slightly different wording and cropped photos becomes one listing.
- Publish. Clean listings go to the client's PostgreSQL database, which the site reads from.
What made it hard
Facebook shows very little to automated visitors and changes its page structure often. A logged-out visitor sees only a group's most recent posts, so we visit each group every two hours and pick up new posts before they scroll out of reach. A daily check compares post counts per group, and a group that suddenly returns nothing is usually the first sign that Facebook changed its markup.
Duplicates were the product problem. If the same flat shows up five times, the site is no better than the groups it replaced.
Result
- About 4,000 listings a month from 40 groups, after duplicates are merged
- A new post reaches the site within two hours
- The site launched in September 2025 with over 3,000 listings already searchable
Tell us what you need
hello@anyx.solutionsSay what you're trying to find out, which websites have it, and where the results need to end up. Rough volume and how often it should run help too. We read every message and reply ourselves.