Methodology
EmlakIQ is built around preserving evidence, normalizing property data, resolving identity, and serving every output with explicit scope and confidence context.
Operating discipline
Pipeline
TKGM, municipal zoning, AFAD, MTA, TUIK, and market feeds are stored raw with file, period, and geometry intact.
immutable raw files · source window · ingest audit
Every batch is sealed with manifest references, load dates, source family, and validation outputs.
manifest:imar_ibb_2026_03 · replay-safe lineage
Address, geometry, parcel, and asset identity are resolved into one schema with type and date discipline.
entity resolution · schema discipline · field contracts
Zoning, risk, market, and coverage fields are promoted with supported / partial / missing logic rather than narrative shortcuts.
known vs inferred vs missing · freshness kept explicit
Property, parcel, district, and period context are joined into one profile for underwriting and screening surfaces.
property + parcel + district + market context
Verify, Analyst, Radar, and API outputs enforce the same evidence contract so product surfaces stay aligned.
same snapshot contract across products
Layer detail
Cadastre, zoning, market, macro, and risk records first enter the system as evidence. Imperfect or incomplete records are documented rather than hidden.
This layer preserves what the raw record actually says. The goal is a source trace that can be replayed and audited later.
Identity fields, geometry, dates, numerics, and text keys are moved into one contract. Ambiguous matches are never presented as settled fact.
Zoning, risk, market activity, and project inventory are merged through domain rules so the final structure can support decisions.
Fields from sources or deterministic transformation stay separate from modeled interpretation. Every summary must show the evidence beneath it.
Reports and API payloads are both tied to the same evidence contract. When the same snapshot is requested again, the same output can be reproduced.
Core principles
Unfilled fields are marked as unknown, inferred, or unsupported. Gaps are never concealed behind polished prose.
Every field ties back to a source family, ingestion path, and timestamp. If the lineage is unclear, the field does not belong in the output.
The same property under the same data state and version should yield the same result. The process is not allowed to drift silently.
Unsupported geographies and data types are surfaced directly rather than omitted. The user sees the limit inside the record.