Transparency is the methodology. Every number in your dashboard either comes from one of the sources below or is derived from them via the formulas in our methodology. We re-verify the inputs every refresh and publish the validation results (median 0.11% error vs. actual published tax bills).
Pennsylvania State Tax Equalization Board
The CLR is the ratio of assessed value to current market value, published annually by the PA STEB. Effective July 1 through June 30. We use the current factor for every gap calculation in our screening.
Montgomery County Treasurer / Board of Assessment
Per-municipality and per-school-district mill rates. Combined with county and MCCC mills to compute each property’s exact total mill rate. We don’t use a county-wide average — the lookup is keyed by city and ZIP, validated against actual published tax bills.
Montgomery County Board of Assessment
Assessed values, property characteristics (sqft, year built, property type, owner of record), and parcel identifiers come directly from the county assessment roll. The full commercial roll is refreshed monthly and screened against the methodology.
Per-county portal — full assessment-history feed
The 48-month no-county-activity guarantee that powers our Premium tier is backed by per-parcel multi-year assessment-history pulls from each county's official portal. For each property we extract every assessment-change row going back as far as the county records (1992+ for Montgomery County). Each row captures the effective date, appraised value, assessed value, and the reason for the change (APPEAL, COURT STIPULATION, REASSESSMENT, ADDITION, DEMOLITION, etc.). We render the county's own assessment-history page as a PNG per property — this is the court-grade evidence in every Premium delivery.
Data-source-driven tier ceilings
We never sell a lead at a tier above what the data supports. Premium tier requires public multi-year assessment-history data from the county's own portal (so we can back the 48-month guarantee). Standard and Economy tiers use lighter data signals. The matrix:
CBRE, Cushman & Wakefield, Colliers
Property-class cap rates, vacancy, opex ratios, and market rent assumptions used in the income approach come from the published quarterly market reports of the major commercial brokers, specific to suburban Philadelphia. Refreshed quarterly.
Licensed commercial data provider
Comparable sales, deed types (for arm’s-length verification), owner mailing addresses, and PropertyShark-style enrichment data come from a licensed third-party commercial real estate dataset. Comps are filtered to ±60% sqft, same canonical property type, sold within 36 months, same or adjacent ZIP.
LoopNet (CoStar Group)
We cross-reference our primary commercial sales feed against LoopNet — the largest U.S. commercial real estate listing platform, owned by CoStar Group. LoopNet listings supply transaction data, asking prices, photos, and broker-reported sale prices for commercial properties. Where LoopNet and our primary feed disagree on a sale price by more than ~10%, the comp is flagged for manual review before it ships in an evidence packet.
Google Maps Platform
Satellite and street-view imagery on each property detail page is loaded on demand from the Google Maps Static and Street View Static APIs. We do not store the imagery; it is fetched per request to authenticated users.
Montgomery County Property Records portal
The County Assessment Record card on each property detail page shows a screenshot of that parcel’s public record page on the MontCo property records portal. We capture and host these for fast access; the underlying data is publicly available at the source.
Anthropic Claude (Haiku 4.5)
The 3–4 paragraph appeal narrative on each property detail page is generated by Anthropic’s Claude Haiku 4.5 from the property’s validated numerical data (assessed value, income approach, comps, mills, gap, savings). The narrative is structured as: (1) over-assessment finding with CLR context, (2) income approach analysis, (3) comparable sales reference, (4) projected savings + filing recommendation.
Cross-checked against actual published tax bills
Every refresh, we ground-truth our predicted tax bill (assessed × total_mills ÷ 1000) against the actual published bill from the source data. Median error across 8,131 verifiable Montgomery County properties: 0.11%. 93% of mapped cities within ±5%; zero cities off by more than 20%. When a city is missing from the lookup, we back-derive the mills from the actual bill data and add it to the table.