What the pilot proved
WorkingPublic listing extraction, Arabic field normalization, filtering and a local SQLite database with observation history.
2 × 5Complete HTTP runs × pages per run
15 / 15Inspected ads matched five primary fields
14Automated tests passed
PassedDatabase integrity + fresh replay check
The five checked fields were title, advertised price, area, source district and source land use. These checks establish extraction fidelity to the website; they do not independently verify the property. The 15 ads were a purposive sample.
Valuation pendingNo bargain claim is justified by this sample.
Broad district averages would blur the location, phase and land-use differences that determine whether two plots are comparable.
0Maximum qualifying comparables under the current strict matching rule
Every retained price still has an assumed total-price basis. Confirm that basis and build a deeper, dated comparison set before activating any discount threshold.