Methodology

How the Lucen Score™ is actually built.

Public. Versioned. Auditable. Every number we publish can be traced to a primary source, a fetch date, and a methodology version. This is the page we publish so you don't have to trust us — you can check.

See the weights Data sources Fair-housing audit
The formula

One formula. Eleven inputs.

The Lucen Score™ is a weighted composite. Each sub-index is itself scored 0–100 from 3–10 input features. The composite is normalized to 0–100 and published with a confidence band reflecting data completeness.

LucenScore = Σ (sub_indexi × weighti), normalized to 0–100.
confidence_band = 2 × (1 + missing/11 + stale/11). Min ±2, max ±10.
Weights (v1.0.0 — residential rental)

Every weight. Every input.

V1 weights are our prior estimate. As outcome data accumulates (claims, defaults, complaints), weights recalibrate. Every change ships with a version bump and appears in the changelog — never silently.

Sub-Index
Weight
Primary Inputs
HPD
15%
Class A/B/C violations · complaints · vacate orders · harassment findings
DOB
12%
DOB violations · OATH/ECB · facade · boiler · elevator history
Safety
12%
NYPD precinct crime rates weighted by proximity and severity
Risk
10%
FEMA flood · EPA air · climate peril composite
Price
10%
Price per sqft vs. building · neighborhood · city
Infrastructure
8%
Hospitals · parks · shopping · essential services
Transportation
8%
Subway/rail proximity · walk + bike scores
Amenities
8%
Laundry · elevator · pets · storage · rooftop · accessibility
Education
7%
School ratings weighted by distance and grade-level relevance
Noise
5%
311 complaint-derived noise model + geospatial sources
Air Quality
5%
EPA AQI · industrial source proximity · traffic pollution
Context-aware scoring. The schedule above is for residential rental reports. Buyer, insurance, and commercial reports use different weight schedules — documented in the full methodology white paper.
Data sources

Every field cites its primary source.

No aggregated opaque feeds. Every data point in every report includes the source name, fetch timestamp, and a freshness indicator. If a source is down, we surface a wider confidence band rather than silently substituting.

Source
Refresh
Used For
NYC HPD Violations API
Daily
HPD sub-index input
NYC DOB Violations API
Daily
DOB sub-index · facade · elevator history
NYC ACRIS
Weekly
Title · deed · lien · mortgage history
NYC PLUTO
Quarterly
Zoning · tax lot · property characteristics
NYC 311 Complaints
Daily
Noise · pest · heat · quality-of-life
NYC DOF Rolling Sales
Weekly
Transaction history · Price sub-index
NYPD CompStat
Weekly
Precinct crime rates · Safety sub-index
FEMA National Risk Index
Quarterly
Flood · heat · hurricane · wind exposure
EPA AirNow
Hourly
Air Quality sub-index
U.S. Census ACS
Annual
Demographic normalization only · never a scoring input
NYC DOE
Annual
School boundaries · Education sub-index
GreatSchools · Walk Score · ClimateCheck
Monthly
Commercial partner supplements
Fair housing

By design, not by policy.

The methodology never uses protected-class inputs. Every quarter, a third-party fair-housing auditor runs disparate-impact tests and publishes the results. Here's what's in and what's out:

Never used as inputs

  • Race · ethnicity
  • National origin
  • Religion
  • Disability
  • Familial status
  • Gender identity · sex
  • Source of income
  • Age (except in senior-specific housing contexts)
  • ZIP code as a scoring factor (used only as a geographic reconciliation key)

Used as inputs

  • Physical building characteristics (age, size, type)
  • Official violation history (HPD, DOB, OATH, ECB)
  • Geographic features (transit, parks, flood zones)
  • Environmental exposure (FEMA, EPA data)
  • Price comparables (building, block, neighborhood)
  • Official school performance data
  • Publicly filed crime statistics at precinct level
Quarterly bias audit. A third-party reviewer tests score distributions across protected-class-correlated demographic profiles. If any proxy variable encodes a protected characteristic beyond a materiality threshold, it's removed and the methodology is versioned. All audit results are published.
Versioning

Semantic. Transparent. Logged.

MAJOR

Change in the list of sub-indices or their definitions. Rare. Requires formal migration notice.

MINOR

Material weight rebalancing (any weight shift >3 percentage points). Published with full changelog.

PATCH

Input-data source changes, normalization tweaks, bug fixes. Logged; no rescore of historical reports.

Every report is stamped with the methodology version used. Historical reports are never retroactively rescored — if you want a fresh score, you request one. This preserves underwriting audit trails for regulated partners.
Disputes

5-day SLA. Every time.

Property owners and residents can challenge a score by submitting evidence (e.g., a resolved violation not yet reflected in HPD). Every dispute is triaged within 5 business days, and valid corrections trigger an immediate rescore. All outcomes are logged for audit.

1

File a dispute

Link on every property report. Attach evidence (official documents, case numbers, resolution letters).

2

Triage

Our data team reviews against the original source. Decision within 5 business days.

3

Correction or response

Valid corrections trigger immediate rescore and a note in the report's revision history.

Read the full white paper.

Every weight derivation, every source, every audit result — in one technical document.

Download v1.0.0 (PDF) Ask a question