Patent Analysis

They are a useful but partial proxy. Patents capture codified, patentable inventions, but the propensity to patent varies greatly across industries (high in pharmaceuticals and electronics, low in software or services), firms, and countries, and much innovation is never patented. Strategic and defensive patenting can also inflate counts. Patent analysis is therefore most reliable for patent-intensive fields and is strongest when triangulated with other indicators such as publications, R&D spending, and market data.

A single invention is often protected by separate patent applications filed in multiple jurisdictions (e.g. the US, Europe, Japan). A patent family groups all these related filings of the same invention. Counting individual documents without consolidating families overstates activity and biases comparisons toward inventions filed in many countries, so serious analysis works at the family level and uses family size itself as an indicator of an invention's perceived value.

Common data sources include the EPO's PATSTAT, Clarivate's Derwent World Patents Index, Google Patents, and Lens.org, alongside national and WIPO collections. Analytical tools range from Derwent Innovation and VantagePoint for cleaning, indicator computation, and mapping to general data-science and text-mining environments. The choice depends on coverage needs, the depth of citation and classification data required, and whether open or commercial access is preferred.