Field Change Order Evidence Tracker — Nkosi Felix, Senior Data Platform Engineer

Problem

Field change requests arrive as notes, photos metadata, and cost-risk estimates from multiple people. Before a change is approved, someone needs to validate required fields, catch duplicate submissions, and keep an evidence trail that survives a dispute.

Data boundary

Synthetic project data only, generated for this lab. No real client, contractor, or project data.

Pipeline

Ingest synthetic change-order record → validate owner, status, and cost-risk fields → detect duplicates → hold for owner approval → produce a review-ready evidence packet.

This pipeline shape mirrors the document-intake and validation-routing pattern in the NBCUniversal experience entry (ACCEPT/QUARANTINE/SYSTEM-failure routing) — connected here only because that pattern is already public evidence, not a new claim.

Validation rules

Sample synthetic change-order records and validation outcomes
Field	Rule	Sample value	Outcome
change_order_id	Required, must be unique	CO-1042	ACCEPT
owner	Required, must match an approved owner list	(missing)	QUARANTINE — missing owner
cost_risk_estimate	Required, numeric	"tbd"	QUARANTINE — non-numeric estimate
submission_hash	Must not duplicate a prior submission	matches CO-1039	QUARANTINE — duplicate detected

Output

A review-ready evidence packet per change order: validation outcome, duplicate status, and an approval boundary that keeps the decision with the project owner.

Failure modes

Missing owner or status field → quarantined, never auto-approved.
Duplicate submission → flagged and held, not silently merged.
Non-numeric cost-risk field → rejected with a structured reason.

What this proves

Document intake, field validation, duplicate detection, and an owner-approval boundary applied to a construction/infrastructure-shaped change-management problem.

What it does not claim

This lab uses synthetic project data only. It does not claim experience with any specific contractor, project, or real construction dispute.