Methodology & Data Sources

Overview

CityLens is a building intelligence platform that aggregates public-record data from New York City and Westchester County agencies to produce a unified view of each tax lot's energy posture, regulatory exposure, transaction history, and financial distress signals. It is a research and prioritization tool — not a certified compliance system, a legal opinion, or a substitute for due diligence on any individual asset.

All scores and grades are derived from publicly available datasets. CityLens does not access private financial records, lease data, or non-public agency files. Refresh cadences are determined by each source agency's publication schedule; CityLens ingests new snapshots within 48 hours of upstream publication.

The platform currently covers 1,114,426 buildings across the five boroughs of New York City and the principal municipalities of Westchester County. Of these, 1,083,437 structures are rendered with detailed 3D geometry from the NYC Department of Information Technology and Telecommunications (DoITT) building footprints dataset, served as OGC 3D Tiles. CityLens hosts 8,746 tile objects on Wasabi S3-compatible storage at https://s3.wasabisys.com/incentedge-3dtiles/tileset.json?v=2 with a Cache-Control: max-age=86400, immutable header. Tax-lot identifiers (BBLs) are sourced from MapPLUTO; approximately 222,000 structures in the DoITT tileset do not match a current PLUTO tax lot (see BIN vs BBL below).

Refresh cadences vary by source: MapPLUTO is published quarterly by DCP and refreshed weekly when DCP posts interim corrections; Local Law 84 and Local Law 97 datasets are ingested quarterly; DOB violation and permit feeds and ACRIS recordings are ingested daily; the tax-lien sale eligibility list is ingested within 48 hours of its annual release; RPIE filings are ingested annually after DOF posts the cycle in November. The Composite Distress Score is recomputed nightly in the current Phase 2 build and is targeted to move to a quarterly canonical refresh once production calibration completes.

CityLens information is provided as-is for informational research purposes. Users should independently verify any signal before acting on it. CityLens does not guarantee the accuracy, completeness, or currency of any data point and is not a substitute for title searches, ALTA surveys, environmental site assessments, certified energy audits, or legal due diligence. Users assume sole responsibility for independent verification before relying on a CityLens signal in any decision.

The 7-Signal Composite Distress Score

The Composite Distress Score is a 0–100 index computed for each building where sufficient data exists. Higher scores indicate greater financial or regulatory distress. The score is designed for portfolio screening — it surfaces buildings that warrant deeper investigation, not buildings that are definitively distressed.

The score is a weighted aggregation of seven independently sourced public-data signals: tax-charge growth, lien-sale notice status, HPD violation counts, LL97 penalty exposure, DOB open-permit posture, ACRIS recording activity, and modeled RPIE income decline. Each signal contributes a bounded subscore; the composite is the sum, scaled to 0–100 and bucketed into four ordinal tiers — low, medium, high, and unknown — for display in the building panel and map color overlays. The unknown tier is reserved for records with insufficient inputs to produce a reliable composite (currently 252,645 BBLs marked tier='unknown' pending Phase 2 score-engine refresh).

The composite is recomputed on a refresh cadence aligned with input-data availability. The current Phase 2 build runs nightly against the most recent input snapshots; the production target is a quarterly canonical recomputation with intra-quarter spot updates when a high-weight input (LL97, tax liens) refreshes. The score is informational research output, not a prediction of default, foreclosure, bankruptcy, regulatory action, or financial distress in any legal sense. CityLens does not represent that any particular score corresponds to any particular outcome for any particular building. Users should treat the score as a starting point for independent due diligence, not as a verdict.

Score weights are calibrated against a backtested set of buildings that experienced documented adverse events (recorded liens, foreclosure filings, sustained code-violation campaigns) between 2019 and 2024. The calibration methodology, weights table, and backtesting confusion matrix are maintained in CityLens internal technical documentation and are available to enterprise customers under NDA. Outside counsel and regulatory reviewers may request the calibration memo through the dispute channel below.

Tax Charges Growing

Input source: NYC Department of Finance (DOF) property tax billing records, accessed via NYC Open Data dataset w2pb-icbu. Weight: Up to 15 points (approximately 15–20% of composite, subject to recalibration). Refresh cadence: Quarterly, aligned with DOF billing cycles (January, April, July, October). Confidence band: High for Class 1 and Class 2 properties; moderate for Class 4 (commercial) due to assessment lag. Known limitations: Assessment challenges and abatements in flight will reduce apparent growth until the DOF posts the final adjusted bill. Buildings with active J-51 or 421-a abatements may show artificially flat charge growth.

The Tax Charges Growing signal measures the rate of increase in unpaid DOF tax balance quarter-over-quarter, computed by the SQL function tax_balance_growing(bbl, quarters=4) against the rolling four-quarter DOF billing snapshots. The function returns a positive value when the unpaid balance has grown in two or more of the prior four quarters, scaled by the growth magnitude relative to the building's assessed value. A building whose unpaid balance has grown for four consecutive quarters scores at the top of the band; a one-time arrearage that is subsequently paid down does not.

This signal carries high confidence because the NYC DOF is the authoritative source for property tax billing and CityLens ingests the published amounts directly without re-derivation. The principal limitation is that the dataset does not distinguish disputed charges, abated charges, or charges subject to active payment plans from legitimate delinquency. A building actively contesting its assessment through Tax Commission proceedings may appear to have growing unpaid charges even when the underlying liability is not yet adjudicated. Users investigating a flagged building should verify the status of any active assessment dispute before drawing conclusions.

Lien Sale Notices

Input source: NYC Department of Finance lien sale eligible property list, published annually. NYC Open Data dataset 9rrd-yx27. Weight: Up to 20 points (lien sale eligibility is a strong distress signal). Refresh cadence: Annual; list published each spring for the July lien sale. Confidence band: High — this is a binary flag derived directly from the DOF eligibility list. Known limitations: Properties can be removed from the list after a payment plan is entered. CityLens reflects the most recent published list; mid-cycle removals may lag up to 30 days.

CityLens has loaded 264,000 lien sale notices spanning the 2019–2025 publication cycles. The NYC lien sale was administratively suspended during the COVID-19 public-health emergency (2022 and 2023 cycles were skipped); the program resumed in 2024. CityLens preserves the historical annual snapshots so users can see whether a building has appeared on multiple consecutive lists or appeared once and was subsequently removed. The most recent published list drives the active flag in the composite; prior years' appearances are visible in the building detail panel as a history strip.

The principal limitation is that lien removal between annual snapshots is not captured in real time. A building that pays down its arrearage or enters a DOF payment plan after the list is published — but before the July lien sale — will not be removed from the CityLens snapshot until the next annual list is published. Users should verify current lien status with NYC DOF directly. The flag indicates the building appeared on the most recent published eligibility list, not that a lien has been sold against the property.

HPD Violations

Input source: NYC Department of Housing Preservation and Development (HPD) building violations, accessed via NYC Open Data dataset wvxf-dwi5. Weight: Up to 25 points, tiered by violation class (Class C Emergency = highest, Class A = lowest). Refresh cadence: Daily from HPD; CityLens ingests nightly. Confidence band: High for open violations; moderate for recently closed violations where HPD's close date may precede CityLens' next ingest. Known limitations: HPD violations are property-level, not unit-level. A single Class C violation on a 500-unit building carries the same raw flag weight as on a 4-unit building. Phase 2 will introduce per-unit normalization.

The HPD violation signal is computed from open Class B (hazardous) and Class C (immediately hazardous) violations published in NYC Open Data dataset wvxf-dwi5. Class A non-hazardous violations contribute a smaller weight, reflecting the lower severity assigned by HPD itself. Confidence is high overall because HPD is the authoritative source for housing-code enforcement and CityLens does not modify the raw counts.

The principal limitation is that violation counts are not currently normalized by building size or unit count. A 100-unit building with 20 open Class B violations may have a better per-unit safety record than a 10-unit building with 5 open Class B violations, but the unnormalized composite weights them roughly equivalently. Phase 2 work will introduce per-unit normalization using PLUTO unit counts; the change will be recorded in the methodology version history when it ships.

LL97 Penalty Exposure

Input source: NYC LL84 energy benchmarking data (annual), used to derive estimated LL97 penalty under the carbon intensity thresholds published by NYC DEP. NYC Open Data dataset ujsc-un8m. Weight: Up to 40 points, scaled by estimated annual penalty dollar amount. Refresh cadence: Annual; LL84 submissions due May 1 each year, posted by NYC by September. Confidence band: Moderate. Penalty estimates are modeled from benchmarked energy use intensities; actual penalties depend on owner-submitted compliance documentation not available in the public record. Known limitations: Buildings that have filed emissions reduction plans or purchased RECs may have lower actual penalties than the model projects. See "Modeled vs Reported Energy Grades" section for grade_source detail.

LL97 penalty exposure is calculated from the actual_emissions column of the LL84 benchmarking dataset compared against the emission_limit_2024 column, which encodes the applicable Local Law 97 emission limits for the 2024–2029 compliance period. The signal models the projected annual penalty dollar amount that would result if reported emissions persisted through the compliance period and no mitigating actions (Renewable Energy Credits, emission-reduction plan filings, retrofits, beneficial electrification) were applied. The score scales with the magnitude of the modeled penalty.

This signal applies only to buildings that meet the LL97 covered-building threshold (generally 25,000 gross square feet or greater, with specific exemptions and inclusions defined in the statute). Buildings below the threshold are not in the LL97 universe and do not generate a penalty-exposure subscore; the LL97 weight is reapportioned across the remaining signals for sub-threshold buildings. CityLens does not estimate LL97 penalties for buildings not covered by the law.

Confidence is moderate because the underlying calculation is a deterministic application of the published DEP emission-limit schedule to reported LL84 data, but the model cannot account for owner actions not reflected in the public record. A building that has filed an emission-reduction plan with DOB, purchased RECs through a compliant pathway, or executed a beneficial electrification retrofit may face an actual penalty materially lower than the modeled exposure. The modeled penalty is a research signal derived from public benchmarking data and the published statutory schedule — not a regulatory enforcement projection.

DOB Open Permits

Input source: NYC Department of Buildings (DOB) permit issuance and status records, NYC Open Data dataset ipu4-2q9a. Weight: Up to 15 points. Refresh cadence: Daily from DOB; CityLens ingests nightly. Confidence band: High for permit count; moderate for permit age (DOB updates closure dates with some lag for older permits). Known limitations: Not all open permits indicate distress — a large renovation permit is a positive signal, not a negative one. The current model does not distinguish between permit types. Phase 2 will weight by permit category (new building vs. alteration vs. work without permit).

The DOB Open Permits signal counts open permits associated with a BBL and currently applies a uniform weight regardless of type. The signal is intentionally low-weight (roughly 5% of the composite in current calibration) because open permits are ambiguous: an open Alt-1 alteration permit may indicate an active improvement program (positive operational signal), while an open work-without-permit complaint or stalled construction job may indicate operational difficulty (negative signal). The current model does not distinguish; it counts.

Phase 2 work will introduce per-job_type weighting so that work-without-permit and stalled jobs contribute negatively while active alteration and equipment-installation permits contribute neutrally or positively. Until then, users investigating a flagged building should review the underlying permit list in the building detail panel to determine whether open permits represent improvement activity or operational difficulty. The change will be recorded in the methodology version history when it ships.

ACRIS Activity

Input source: NYC Office of the City Register ACRIS (Automated City Register Information System), transaction and lien document feed. NYC Open Data dataset 8h5j-fqxa. Weight: Up to 10 points (low recent activity is a mild distress signal for large assets). Refresh cadence: Daily from ACRIS; CityLens ingests nightly. Confidence band: Moderate. ACRIS document recording dates lag the actual transaction closing by days to weeks. Known limitations: ACRIS activity is not normalized for asset size or type. A Class A office building transacting every 10 years is normal; a small multifamily building with no transaction in 20 years is a stronger signal. Current model uses flat thresholds; Phase 2 will introduce peer-group normalization.

The ACRIS Activity signal measures the frequency of recorded deeds, mortgages, mortgage assignments, satisfactions, and lien filings over the trailing 24 months for each BBL. It contributes positively when activity is unusually high (potential distress-flip pattern: rapid-succession recordings, mortgage assignments to specialized debt servicers, repeated lien filings) and mildly when activity is unusually low for a building of its size and class.

Confidence is high because ACRIS is the authoritative source for recorded instruments in New York City, but moderate at the building level because the signal is not currently normalized by peer asset class — an active institutional investor portfolio may register the same pattern as a distressed-flip portfolio. Phase 2 work will introduce peer-group normalization by PLUTO building class and assessed value range. Users should review the underlying ACRIS document list (deep-linked from the building detail panel) before drawing conclusions about transactional intent.

RPIE Income Decline

Input source: NYC Department of Finance Real Property Income and Expense (RPIE) filings, available for income-producing properties with assessed value over $40,000. NYC Open Data dataset uifc-1ynx. Weight: Up to 20 points. Refresh cadence: Annual; RPIE filings due June 1, posted by DOF by November. Confidence band: Low to moderate. RPIE filings are owner-submitted and subject to amendment. Gross income figures may not reflect net operating income. Known limitations: RPIE is not required for owner-occupied buildings, cooperatives, or condominiums. Roughly 40% of NYC tax lots have no RPIE data. Where RPIE is absent, this signal contributes 0 points and the score is flagged as "Partial Data."

The RPIE Income Decline signal models a trailing-twelve-month NOI estimate for commercial buildings over 40,000 gross square feet and compares it against the prior-year RPIE filing to detect year-over-year decline. Because per-BBL RPIE income figures are anonymized in the public dataset, CityLens reverse-engineers the modeled NOI from DOF assessed-value movements, which themselves incorporate RPIE inputs through the DOF capitalization methodology.

Confidence is explicitly low (model confidence flag = 0.20) and CityLens surfaces this directly in the building detail panel. The signal is included because trailing income decline is a leading indicator of operational distress for income-producing commercial real estate, but should be treated as directional rather than precise. Principal limitations: amendment lag of 12–18 months in DOF's assessed-value response to RPIE filings; absence of RPIE entirely for owner-occupied buildings and most residential cooperatives and condominiums; and the inability of assessed-value reverse-engineering to distinguish income decline driven by vacancy versus rent concessions versus operating cost growth. Users should treat the RPIE signal as a hypothesis to verify, not a finding.

Data Sources

The following table lists all primary data sources ingested by CityLens. "Last Refresh" values are placeholders updated at ingest time by the pipeline.

Most NYC Open Data datasets are published under the NYC Open Data Terms of Use, which permit broad re-use, re-distribution, and derivative-work creation provided attribution to the publishing agency is preserved. Specific terms should be verified at the source URL for each entry; CityLens has reviewed active terms as of the most recent ingest but does not represent that those terms remain unchanged. FEMA NFHL data is a federal work in the public domain.

License flow-through disclaimer. CityLens is a derivative work combining, transforming, and aggregating the source datasets above. Source license terms flow through to derivative use; CityLens does not extend additional re-distribution rights beyond what the source grants. Users who re-distribute CityLens-derived outputs (CSV exports, API responses, screenshots) should review the underlying source licenses to confirm the intended use is permitted. Where an output combines multiple sources, the most restrictive applicable source license governs the combined output. Westchester County data sources are under active onboarding; the same flow-through principle will apply when they ship.

Quarantine and Data Quality

CityLens applies two active quarantine rules that were identified during the initial data import and documented as quality issues P2-DQ-02 and P2-DQ-03.

P2-DQ-02 — LL84 Unit Ceiling Anomalies: The LL84 dataset contains a small number of records where reported gross floor area, unit count, or aggregate emissions fall outside the physically plausible range. CityLens removes from energy-grade computation any LL84 row where total_ghg_emissions > 200,000 MT CO2e, site_eui > 5,000 kBtu/sqft, or source_eui > 15,000 kBtu/sqft. These ceilings are calibrated against the maximum credible values for any single building in the NYC stock, including the largest hospital campuses and trading-floor office buildings. The most consequential record corrected was a benchmarking submission for St. Barnabas Hospital that, if accepted at face value, would have generated a modeled LL97 penalty of approximately $513 million per year. As of the 2025-09 ingest, 1,847 LL84 records are quarantined under this rule. Quarantined buildings display "Data Under Review" in the energy grade field.

P2-DQ-03 — GHG Intensity Check: A secondary rule removes records where the ratio of reported GHG emissions to gross floor area exceeds 0.5 metric tons CO2e per square foot — roughly two orders of magnitude above the worst credible real-world performance for any NYC building type, reliably the result of unit-of-measure errors in the source submission (metric tons reported in pounds, or per-floor figures reported as aggregate). The most consequential record corrected was a Columbia University building submission that would have generated a modeled LL97 penalty of approximately $22.5 million per year. As of the 2025-09 ingest, 312 records are quarantined under this rule. Affected buildings are flagged with "Penalty Estimate Unreliable."

Why quarantine rather than display. Quarantine is the responsible choice for records whose face-value display would mislead users into investigating non-existent regulatory exposure. Displaying an obviously-wrong $513 million penalty would damage trust in the platform and could prompt downstream actions (lender notifications, owner outreach, regulatory inquiries) based on what is in fact a data-entry error in a source dataset. Quarantine preserves the record for audit while preventing the misleading display.

Quarantine is preservation, not deletion. All quarantined rows are retained in the energy_benchmarks_quarantine table with a populated quarantine_reason field. They are excluded from score computation, energy-grade display, and portfolio analytics, but remain visible in the building detail panel under a "Quarantined Records" disclosure. Owners and authorized representatives who believe their data has been incorrectly quarantined may appeal through the dispute channel at methodology@incentedge.com. On successful appeal, CityLens will either re-include the record in active computation or work with the owner to coordinate a corrected submission to the source agency.

Modeled vs Reported Energy Grades

CityLens energy grades derive from two distinct sources, distinguished by the grade_source field returned in the building API response and displayed in the building detail panel.

Reported (grade_source: 'reported'): The energy grade is taken directly from the building owner's LL84 annual benchmarking submission to the NYC Mayor's Office of Climate and Environmental Justice. These grades reflect the owner's own energy metering data and carry the highest confidence. CityLens has loaded 59,713 reported energy grades corresponding to the subset of NYC buildings within the Local Law 33 / Local Law 84 reporting universe. LL84 reporting is generally required only for buildings of 25,000 gross square feet or larger, plus certain smaller buildings on tax lots that meet aggregate thresholds — approximately 5% of the total NYC building stock.

Modeled (grade_source: 'modeled'): For the remaining approximately 1.06 million buildings not in the LL84 reporting universe, CityLens applies a heuristic model. The model uses building class (from PLUTO), year built, gross floor area, borough, and the borough-median EUI for that building class. The output is a letter grade (A–F) presented on the same scale as reported grades but distinguished visually and through the grade_source field. The model is calibrated against the reported-grade universe — for buildings with both a reported and modeled grade, the model matches the reported grade in roughly two-thirds of cases and is no more than one letter off in the remainder. It is not a substitute for an actual benchmarking submission.

Confidence tiers. CityLens publishes a confidence score with each grade. Reported = 1.00. Modeled = 0.40 by default, falling to 0.25 (grade_source: 'modeled_class') where the output is derived from class-level defaults alone (buildings with sparse PLUTO records, missing year-built, or unusual class codes). Buildings for which no grade can be computed carry confidence = 0.10 and grade_source: 'unknown'. The user-facing interface labels modeled grades with a distinct visual treatment (italic letter, lighter color band) and a tooltip explaining the modeled basis; CityLens never displays a modeled grade as if it were reported.

Sample heuristic outputs: a 1925 Class C5 Manhattan walk-up at 8,500 sqft typically receives D (Manhattan walk-up median EUI ~120 kBtu/sqft); a 1990 Class K4 Queens retail building at 18,000 sqft typically receives C; a 2010 Class R4 Brooklyn condo at 22,000 sqft typically receives B. These reflect borough-class medians and will diverge from any individual building's actual measured performance; they are starting points for investigation, not certifications.

BIN vs BBL

The NYC Department of Buildings assigns a Building Identification Number (BIN) to each physical structure. The NYC Department of Finance assigns a Borough-Block-Lot (BBL) number to each tax lot. The relationship is many-to-many: a single tax lot may contain multiple buildings (campuses, multi-building condo developments, parcels with accessory structures), and a building may technically straddle a lot line. Most CityLens analytics are BBL-keyed because tax, ownership, energy, and regulatory data are predominantly published at the BBL level; DOB violations, permits, and the underlying 3D Tiles geometry are BIN-keyed.

Approximately 222,000 structures in the CityLens database have a BIN in the DoITT footprints dataset but do not match any current PLUTO tax lot. The principal reasons: campus subbuildings (university, hospital, large school complexes where the campus is a single BBL but houses many distinct buildings); condo air-rights splits (physical structure pre-dates a subdivision that PLUTO has not caught up to); sub-25,000-square-foot accessory structures (garages, vaults, mechanical buildings, sheds not separately assigned BBLs); and structures on city-owned land where no individual tax lot was ever recorded. These structures appear with the BinOnlyBadge component (amber, "Limited Data") and are excluded from the BBL-keyed Composite Distress Score.

For BIN-only structures, CityLens surfaces what is available: the 3D geometry from the DoITT tileset, the BIN identifier (clickable through to the NYC DOB BIS lookup), and where possible, parent-lot information via the internal bin_to_bbl lookup table maintained from the DCP PLUTO dataset and the DOB BIS database. The lookup is refreshed quarterly. Where a BIN maps to multiple BBLs, the BBL with the larger assessed value is used as the canonical parent for display; users should treat the parent-lot inference as a hint and verify through DOB BIS before relying on the relationship.

Surfacing these 222,000 structures rather than hiding them is a deliberate choice. Hidden structures would create visible gaps in the cityscape (lots that clearly contain buildings rendering as empty), which would damage trust in the platform's completeness and could be misleading in its own right. The "Limited Data" badge communicates the platform's known limitations directly to the user.

Disputing a Score

CityLens scores are derived from public-record data. Errors in the underlying agency datasets — miskeyed BBLs, unreported violation closures, stale benchmarking submissions, paid-down liens not yet reflected, withdrawn DOB complaints not yet posted — can produce inaccurate CityLens outputs. CityLens maintains a formal dispute channel for owners, asset managers, property managers, and authorized representatives who believe a score, grade, or signal is incorrect.

Formal intake channel. Disputes should be submitted by email to methodology@incentedge.com with the subject line "CityLens Score Dispute BBL [bbl]" (substituting the actual borough-block-lot number of the building in question). The email body should include:

1. BBL of the building in question (visible in the CityLens building detail panel and in the URL).
2. Specific score, grade, or signal disputed (e.g., "Composite Distress Score = 78", "Energy Grade = D", "Lien Sale Notice flag = true").
3. The specific underlying signal the dispute targets (e.g., "the HPD violation count includes three violations that were closed by HPD on March 12, 2026 but appear to still be counted as open in CityLens").
4. Supporting evidence documenting the dispute (e.g., an updated LL84 filing receipt, a paid-lien certificate from NYC DOF, a withdrawn-violation notice from HPD, a corrected ACRIS recording).
5. Submitter identification (name, role, authority to act for the building owner if applicable).

A pre-populated email link is available from the methodology section of every CityLens building detail panel.

Service-level commitment. CityLens acknowledges receipt of a dispute within 5 business days and provides a written determination within 30 business days. The determination explains the disposition: accepted with the underlying record corrected (published in the next nightly refresh with a row added to the version history), declined with a stated reason (typically that the dispute targets the agency's published data rather than a CityLens computation, in which case the submitter is directed to the appropriate agency channel), or partially accepted with a stated scope.

Scope of CityLens authority. CityLens does not adjudicate the underlying regulatory finding. If the dispute is that an HPD violation was wrongly issued, the proper channel is HPD's own dismissal process; CityLens cannot remove a violation from its display until HPD posts the closure to the source dataset. If the dispute is that an LL84 submission was incorrectly filed by a prior owner or managing agent, the proper channel is a corrected re-submission to the Mayor's Office of Climate and Environmental Justice; CityLens will quarantine the disputed record and ingest the correction when published.

What CityLens does adjudicate is its own derived computation: weight application, signal inclusion, quarantine determinations, and the correctness of cross-dataset joins (BIN-to-BBL, parcel-to-owner). Where evidence demonstrates an error in a CityLens computation or its interpretation of source data, the score is recalculated and a corrected record is published. Recalculated scores include a correction_log reference visible to authorized portfolio owners.

Version History

The canonical changelog for this document is the git history of src/content/citylens/methodology.md in the public IncentEdge repository. To view the full history of changes — data sources added, quarantine rules introduced, signal weights adjusted, dispute procedures modified — run:

git log src/content/citylens/methodology.md

against a local clone or browse the file's history on the repository web interface. Significant policy changes affecting score interpretation, dispute SLA, or quarantine rules trigger a version-bump tag (methodology-v* series, e.g. methodology-v2.0.0) and a row in the changelog table rendered on this page. Routine content edits, typo corrections, and source-cadence label updates do not trigger a version bump and are visible only in the underlying git history.

Phase 2 changelog rendering. A future iteration will render a human-readable changelog table by parsing git tags matching methodology-v*. Until then, the authoritative changelog is the git tag list and commit history of the source file. Stakeholders requiring a point-in-time citation may reference a specific commit SHA or tag; CityLens commits to preserving the full git history of this file as a public-record artifact.

---

Last reviewed against current data: 2026-05-17. Page generated from src/content/citylens/methodology.md at the commit SHA shown in the footer.