Private by default
We never ask for a name, email, contractor name, full ZIP code or exact property address.
Real projects. Reviewed data. No made-up examples.
Estimates are useful. Completed-project evidence is better. This growing database keeps quotes separate from final costs and releases only privacy-safe distributions whose complete eligible membership matches a fingerprint-pinned reviewed registry release.
Reviewed before use
Editors check arithmetic, completeness, plausibility, date and project scope before approval.
Aggregates, not dossiers
Individual reports remain private. Coverage checks begin at 5 nationally by project, 8 by state + project, 10 by metro + project and 12 by ZIP3 + project; cost distributions use a stricter 12-report floor. These counts create candidates only and never authorize release.
Confidence stays visible
Self-reported, receipt-reviewed and contractor-verified inputs are not treated as identical evidence.
RELEASED, ANONYMIZED COHORTS
What the community dataset can honestly show
Completed-project distributions appear only after editorial review, the necessary publication floor and a fingerprint-pinned reviewed registry release. Quote-only records stay outside actual-cost percentiles.
Checking released aggregate groups…
No figures are shown until the dataset response is verified.
PUBLIC RELEASE RULES
Approval and threshold depth do not publish a report.
A row can contribute only when it describes a completed final cost, carries the current privacy acknowledgement, passes editorial approval, reaches quality 0.50 and stays at or below outlier score 0.65. Quote-only and legacy-unclassified rows never enter completed-cost distributions. Even after the floor is met, the complete eligible cohort must exactly match the count, review cutoff and membership fingerprint in a reviewed release registry entry. Individual rows, exact ZIP codes, street addresses and contributor identities are never published.
- 5
- National project minimum
- 8
- State + project minimum
- 10
- Metro + project minimum
- 12
- ZIP3 + project minimum
HOW THIS BECOMES INTELLIGENCE
One report is a clue. A reviewed dataset is a model.
- Validate
Check final and change-order costs, component arithmetic, project month, ZIP3 and obvious duplicates.
- Score
Assign a data-quality and verification level without pretending a recollection equals an invoice.
- Aggregate
Form a P25, median and P75 candidate only after at least 12 eligible completed projects clear the cohort gate. Publish only after a separate review pins the complete cohort membership and fingerprint in the release registry.
- Improve
Use seasonal, regional, labor and material patterns to recalibrate future estimates.
