Methodology
Short, honest notes on provenance, validation, and the limits of the current extract.
What this is
SerffSurf is a workers' compensation (and expanding multiline) rate-filing intelligence product built from public SERFF filings, 2020–2026. Coverage spans both NCCI and independent-bureau states and grows as the acquisition pipeline works through remaining U.S. jurisdictions.
It is a market-intelligence terminal for pricing teams — not a filing repository and not actuarial advice.
How numbers get here
Spine parse (deterministic): the main SERFF “PDF Pipeline” document for each filing is parsed into Filing-at-a-Glance fields, company tables, section text, and correspondence letters.
Entity resolution: company names map to NAIC cocodes and groups where possible; rate rows join through a resolved company mart.
LLM extraction (second pass): LCM and deviation facts are extracted only from supplied filing context. Every accepted fact must include a verbatim evidence quote that appears (alphanumeric-normalized) in that context.
Two-tier reasonability gates sit on top of the hard bounds: values inside typical actuarial ranges are green; values outside typical but still inside a wider allowed range are amber — valid and shown, but flagged for expert review. Values outside the allowed bounds are hard-rejected and never surface on this site.
Hard validity gates: LCM values outside 0.4–6.0 and deviation percentages outside −90…+300 are marked invalid. Invalid facts are retained with error metadata; this site only surfaces valid (green and amber) facts.
Comparing LCMs across states
A loss cost multiplier is applied to the loss-cost basis that state uses. NCCI advisory loss costs and each independent bureau's pure premiums sit on different bases, so LCM levels are not apples-to-apples across those regimes — a higher median LCM in an NCCI state than in an independent-bureau state can be a basis difference, not proof that the NCCI state is more expensive.
Within one bureau or one state, LCM levels and trends are comparable. Prefer within-regime or within-state reads when ranking carriers or markets.
What to trust — and verify
Evidence quotes are the credibility layer. Prefer them over bare numbers when a figure will drive a decision.
Company-table rate rows are generally more reliable than filing-overall figures repeated across affiliates.
Table-jumble exhibits can yield verbatim-true but semantically mislabeled facts. A hand-labeled gold sample is the planned precision check before treating extracts as benchmark-grade.
Always verify material figures against the underlying SERFF filing.
Architecture of this site
Authenticated product routes read from a Postgres slice API (/api/v1). There is no static /data/*.json corpus — wholesale JSON downloads are blocked at the edge.
MCP tools share the same Postgres data plane (computed analytics + cited semantic search). Aggregate-only partner codes never receive carrier-identified rows.