Transparency about our own process is as important as the transparency we demand from government. Here's exactly how we work.

Data Collection

We run automated scrapers against 15+ public government data sources on a regular schedule. These include county and town budget documents, meeting agendas and minutes, New York Board of Elections campaign finance filings, NYSED school district reports, vendor contract listings, and county legislature records.

All scrapers include change detection — we only process documents when they've actually been updated. Raw source documents are always preserved for audit.

Cross-Referencing

Our database links entities across datasets using fuzzy matching. This allows us to answer questions like: "Which campaign donors also have county vendor contracts?" or "How does this town's budget growth compare to inflation and peer municipalities?"

When we identify a pattern — such as a donor-vendor overlap — we report the connection and the dollar amounts, not a conclusion about intent. Correlation is documented. Causation requires more evidence.

Verification Standards

Every number published must be traceable to a specific page in a public record. We cite our sources at the end of every Evidence Brief and link to original documents when available online.

When data appears anomalous, we verify against multiple sources before publishing. If a number seems wrong but we can't resolve the discrepancy, we note the uncertainty explicitly.

Comparisons

A number without context is noise. We benchmark Rockland County data against peer counties, state averages, and national standards where available. When we say spending is "above average," we show you the comparison set.

Corrections

We will get things wrong. When we do, we correct the record promptly and transparently. Corrections are noted at the top of the affected article and logged on our corrections page.

If you spot an error, please let us know. Accuracy matters more than speed.

AI Disclosure

We use AI tools in our data processing pipeline — for extracting structured data from PDFs, normalizing records across different formats, and generating initial draft analyses. All AI-assisted content is reviewed and edited by humans before publication. Our daily podcast uses AI narration.

The analysis and editorial judgment are human. The scale is AI-assisted. We believe this is the only way a small team can provide the breadth of coverage that 330,000 residents deserve.