Our Methodology
ClaimBack is built on the principle that every statistic we display must be traceable to a primary, publicly-citable source. This page explains how we collect, normalize, and validate data — and what we do when data is unavailable.
1. Source Selection
We only use primary public sources: government agencies (CMS, NAIC, APRA, MAS), statutory ombudsmen (Financial Ombudsman Service, AFCA, FIDReC, OLHI), and independent non-profit research organizations (KFF). We do not use insurer self-reported marketing statistics, broker estimates, or media-cited figures without tracing them to a primary source.
2. Data Collection
Data is collected from publicly-available PDFs, data portals, and structured datasets. Where sources publish data as annual reports (e.g. FOS, AFCA), we extract the relevant figure, note the exact report edition and page reference, and store it with the reporting period. Where sources provide live lookup tools (e.g. NAIC CIS), we link directly to the tool rather than cache potentially stale figures.
3. Normalization
Each source uses different terminology and reporting formats. We normalize data into a common schema: geography (country/state), insurer name, denial rate (as a decimal proportion), appeal overturn rate (as a decimal proportion), complaint count, complaint ratio, source ID, citation label, and reporting period. Fields not available in the source are stored as null — never estimated or inferred.
4. Missing Data Handling
When a data point is not available from a primary source, we display "Not available from current source" or simply hide the data block, rather than showing an estimate. We do not use placeholder values that could be mistaken for real statistics. If we believe important data exists but we cannot locate the primary source, we note this explicitly.
5. Update Frequency
Most of our sources publish annually. We aim to refresh data within 90 days of a source's annual publication. Each data record stores a last_fetched date. Pages display the reporting period of the underlying source alongside any statistic, so readers always know how current the data is.
6. What We Do Not Do
We do not: extrapolate from small samples to national figures; adjust or weight published figures; aggregate incompatible metrics across jurisdictions to produce a single "global" figure; or republish insurer-disputed statistics without noting the dispute. We also do not claim our denial letter tool or appeal guidance constitutes legal advice.