Methodology

Sources

• SAHA İstanbul — 1,213 firms
• OSSA / OSTİM Savunma — 306 firms
• BASDEC — 96 firms
• HUKD — 48 firms
• DASSAD — 19 firms
• SİBER KÜME — 216 firms

Deduplication

Across the six clusters, raw records total 1,898. Cross-cluster duplicates are merged using a two-pass pipeline:

1. Domain match: firms sharing a normalized website domain (eTLD+1) are treated as the same entity.
2. Fuzzy name match: firms with very similar names (≥95 token-set similarity) without a domain conflict are merged.

Merged result: 1,751 unique firms. 13 borderline pairs are held in a manual review queue.

Geocoding

Addresses are geocoded with OpenStreetMap Nominatim. Street-level coordinates are used when available; otherwise the firm is placed at the province centroid.

Known limitations

• TSSK (Ankara / ODTÜ Teknokent, ~114 firms) was excluded — its public directory was returning 404 at scrape time.