Replicable Patent Indicators Using the Google Patents Public Datasets

How can we standardize and replicate patent indicators using the Google Patents Public Datasets?

As patent data become more accessible, there is a pressing need for standardized methods to analyze these data. Inconsistencies and quality issues in patent reporting can lead to flawed conclusions about patent value and technological impact. Establishing reliable and replicable patent indicators is essential for advancing patent analytics and providing valuable insights to researchers and practitioners.

We constructed a variety of patent indicators using the Google Patents Public Datasets, including citation counts and measures of patent originality. We designed and provided BigQuery code through an accessible Kaggle notebook, illustrating how to efficiently extract and analyze patent metrics. The sample consisted of 1600 US patents granted between 2002 and 2015, selected from four biotechnology IPC sub-classes, each with 400 randomly chosen patents.

Our analysis revealed several key metrics: the count of forward citations increased with longer time windows, demonstrating trends in citation practices across IPC sections; the originality indicator clarified how patents build upon diverse technological backgrounds; and a majority of patents exhibited originality scores close to 0, indicating that many inventions do not significantly innovate beyond existing patents. These findings underscore the importance and variability of patent metrics and provide a foundation for future research in the field.

• Sample includes 1600 US patents granted between 2002 and 2015, selected from four biotechnology IPC sub-classes (A61K, G01N, C12P, C07K).
• Data are sourced from Google Patents Public Datasets.
• Replication code is available in a Kaggle notebook.

• The indicators are constructed based on the assumption that citation counts reflect the technological importance of a patent.
• Inventive originality is assumed to correlate with the diversity of technological backgrounds of cited patents.
• The metrics assume that patent data from different IPC sections and time windows can be compared meaningfully.