Every observation in Landvex passes through a multi-stage verification chain before it influences a score. Confidence is earned, not assumed.
Every data point undergoes sequential verification. A failure at any stage removes the observation from the scoring pool.
Each contributor is identity-verified before any observations are accepted. Device binding, location history, and activity patterns are monitored continuously for anomalies.
GPS coordinates are validated against declared location, movement patterns, and historical data. Spoofed, interpolated, or implausible locations are automatically rejected.
Computer vision models review submitted images for specification compliance, content authenticity, and data integrity. Manipulated or non-compliant images are flagged and excluded.
Observations are checked against task specifications: correct asset type, required angles, mandatory data fields, and temporal constraints. Incomplete submissions are rejected, not approximated.
Edge cases, borderline confidence scores, and flagged observations are reviewed by trained human analysts before entering the scoring pipeline. Automated decisions are not final.
Published targets create accountability. These are the standards Landvex commits to across every data collection mission.
Landvex attaches a 0–100 confidence score to every data point. Scores are transparent, composable, and auditable.
Example: Norrmalm commercial vacancy observation, June 2026. Scored across four independent factors.
When observed data conflicts with official sources by more than 2 standard deviations, the observation is automatically flagged for review rather than silently averaged away.
Any observation diverging from official sources by more than 2 standard deviations triggers a structured review process before inclusion in scores.
Landvex stores observations with a complete provenance chain. Nothing is anonymised away if it’s needed for audit.
Each observation is linked to a hashed contributor identity. Allows quality tracking without exposing personal data in standard outputs.
Precise capture timestamp stored with every observation. Temporal accuracy is critical for decay weighting and freshness scoring.
Device model, OS version, and sensor data retained for quality auditing and anomaly detection. Used internally; not surfaced in standard reports.
Full GPS trace stored per observation session. Enables movement-pattern validation and retroactive quality review if anomalies are detected.
Standard Landvex output includes final scores and contradiction flags. Enterprise clients can go deeper.
Enterprise clients can access raw confidence scores via API, review individual observation-level provenance, and receive structured audit trails for regulatory or investment committee use.
Whether you need technical documentation, API access, or audit trails — get in touch and we’ll respond within one business day.