In the past, it was challenging to rank and evaluate patent attorneys, law firms, or patent portfolios using quantitative measures. While previous reports have ranked firms or companies based on their filing activity, little effort has been made to evaluate the quality of an attorney’s work or a law firm’s performance, or to evaluate the quality of a patent portfolio. This was due to the unavailability of data from the patent office and the high cost and complexity of processing large data sets.
The landscape has changed with advancements in computational power and increased data availability from sources such as the USPTO, the PTAB, and PACER for patent litigation. These developments have enabled more comprehensive evaluations, making it possible to assess not only activity but also performance and impact within the patent system.
Entity Name Resolution
Since 2016, we have been analyzing patent data to rank law firms and companies based on their activity. One of the biggest challenges in this process has been data cleaning. Law firms and companies often file under multiple names, including subsidiaries, while mergers and acquisitions can significantly impact rankings in ways that do not reflect organic growth. Additionally, clerical errors in filings further complicate identifying the correct entity names.
To overcome these challenges, the Patexia Data Science team has spent the past years refining its Entity Name Resolution algorithm. This advanced system enables us to automatically extract and verify key information from thousands of PDF documents, including the names of attorneys handling specific patent applications. While no algorithm is flawless, continuous improvements have significantly enhanced its accuracy. By leveraging Natural Language Processing (NLP) and Machine Learning techniques, we continuously train and refine our system, ensuring greater data integrity and reliability in our rankings.
Interviews as an Activity Indicator
Building on our previous ranking methodologies, we continue refining our approach to evaluating patent attorney activity. Last year, we introduced examiner interviews as an additional indicator of activity, recognizing their crucial role in the patent prosecution process. Interviews between patent examiners and applicants or their representatives serve as key touchpoints in the prosecution process. These discussions, conducted in person, by phone, or via video conference, help resolve objections and clarify claim scope, ultimately shaping the final outcome of a patent application. While traditionally, the number of issued patents and filed applications served as primary activity metrics, our analysis has confirmed that interviews provide valuable insight into an attorney’s engagement and influence in patent prosecution. By refining our ranking model to incorporate interview data more effectively, we continue our commitment to delivering the most accurate and comprehensive assessment of patent attorney activity. These improvements ensure that our rankings reflect not only filing volume but also the strategic involvement of attorneys in securing strong patent protection for their clients.
Defining and Measuring Important Qualities
Aside from a clean data set, we had to define factors that were important from the perspective of different stakeholders (i.e., Corporations (Assignees), Law Firms, and Examiners). We came up with the following five high-level factors to measure the overall performance of any stakeholder:
- Activity – The number of issued patents and filed applications by a corporation and law firm, or the number of issued patents and interviews participated by attorneys
- Success – Obtaining a patent or patent claim from the perspective of the applicant, law firm, or attorney
- Quality – The claim quality, broadness, novelty, and market coverage that ultimately make a patent or a claim valuable.
- Efficiency – The speed and simplicity of the process of obtaining a patent
- Cost – The overall cost, including attorney’s fee and the USPTO fees, for obtaining and maintaining a patent
We then defined the performance of each of the stakeholders as a weighted average of those five factors. For example, we define the performance of law firms and companies as:
Performance (law firm / attorney) = a1. Success + a2 . Efficiency + a3 . Activity
Performance (company) = b1 . Success + b2 . Activity + b3 . Quality
In our ranking methodology, we distinguished the performance criteria between companies and law firms since the latter are typically not accountable for the quality of the invention itself, including the field, market, and complexity, among other factors. Their objective is to acquire a valid patent with the broadest claim feasible in a cost-effective and efficient manner. However, they are not the ones generating the invention.
On the other side, companies strive to secure the highest quality patents while also minimizing expenses. Although they have less control over the efficiency of the process, they can influence the quality of their inventions.
Important Signals
The Patexia Data Science team extracted more than 20 signals from a variety of data sources and constructed all of the above factors (e.g., success, quality, efficiency, activity) by combining different signals with different weights. For the purpose of this report, these are some of the key signals we used in our analysis:
- Filing Activity Count – A count of the number of issued patents per organization, attorney, examiner, etc.
- Interview Count – Number of interviews in which the patent attorney or agent participated during the patent prosecution process
- Allowance Rate (Application-Level) – The ratio of patents issued to applications filed for a certain period of time.
- Allowance Rate (Claim-Level) – The ratio of the claims allowed to claims applied for in the original application.
- Number of Office Actions – Average number of Office Actions during the examination process.
- Pendency – The number of days from filing date to issuance or abandonment of an application.
- Number of Time Extensions – The total number of times the attorney or law firm requested an extension of time to respond to an Office Action. Some of these extension requests may require additional fees of up to $3,000 for large entities, which often means a delay in the examination.
- Ratio of Average Number of Words in Independent Claims (patent to application) – The patent-to-application ratio of the average number of words in independent claims shows how much the claim changed from its original form. A number smaller than one shows that more limitations were added during the examination, which usually limits the coverage of the claim.
Some other factors that we considered for quality measurement were, for example, Forward and Backward Citation, Number of Words in Independent Claims. We are constantly adding new signals to improve our measurements and make them more relevant.
Normalization and Noise Removal
There are several players that could impact the outcome of patent examination:
- Company – Assignees or law firm clients (companies) may have different strategies when it comes to filing. These strategies could influence the work an attorney performs.
- Examiner – While the patent office provides a very detailed guideline for the examination process, each examiner may still act somewhat differently based on his/her judgment and understanding of the invention. The examiner’s assessment of an application could change the final outcome.
- Subject Matter – The field of invention, its complexity, novelty, etc., could change the outcome. For example, some fields are more mature (e.g., agriculture or civil), where inventions tend to be narrower. Some other fields are newer (e.g., IoT, drones). The amount of prior art obviously varies significantly based on the area of invention, which could change the examination time, claim broadness, and many other factors.
- Law Firms / Attorneys – As the person or organization in charge of patent prosecution, law firms and attorneys undoubtedly have a huge impact on the outcome. Depending on which firm takes over the patent prosecution – and even within the firm, depending on which attorney is in charge – we could see substantially different results.
To ensure fairness and remove the impact of external factors such as subject matter, examiner assignments, or company-specific strategies, we refined our methodology to make performance measurements as objective as possible. This approach is crucial, as attorneys working under similar conditions may experience vastly different outcomes due to corporate policies and strategic approaches. For example, the patent prosecution strategies of companies like Apple and Intel may differ significantly, influencing their attorneys’ results in distinct ways.
Companies take different approaches to patent prosecution—some prioritize obtaining the broadest possible claims, engaging in extensive back-and-forth with the examiner, while others focus on cost efficiency, minimizing Office Actions and limiting examiner correspondence to control expenses. These variations can introduce bias if not accounted for.
With access to an extensive dataset covering millions of applications over five years, we leveraged regression modeling to statistically eliminate the influence of such factors when evaluating specific entities. This ensured that when ranking law firms or attorneys, the results were not skewed by differences in corporate culture, examiner tendencies, or subject matter complexities. Without this refinement, direct comparisons would not accurately reflect actual performance.
Scope of Analysis and Timeline
When evaluating patent activity, a five-year window provides a more comprehensive view than a single year. This approach accounts for the average time of about three years required to obtain a patent, allowing for the identification of broader trends and patterns beyond mere activity rankings. Additionally, it minimizes the impact of incomplete data for a specific organization during part of a year. In this report, we analyzed approximately 1.8 million patents and 2.2 million utility applications issued or abandoned between January 1, 2020, and December 31, 2024. To address the lower volume of biotech companies’ patent activity compared to high-tech, we created separate metrics for these categories. Thus, the report is primarily structured around three categories.
- High-tech patents
- Biotech patents
- All patents (everything including high-tech and biotech but excluding design patents)
Since Design Patents, in general, have a different examination and prosecution process, and because a shorter amount of time is required to obtain a design patent, we decided to exclude them from most of this analysis. This helps us focus mainly on utility patents, which are the most popular type of patent.
To keep things simple, we divided the high-tech and biotech patents, using the Technology Center associated with each patent application by the USPTO.
Moreover, we limited our measurements to the top 2,000 most active law firms in different categories (i.e., high-tech, biotech, and overall). That means if, for example, a firm was ranked 2,200th in terms of Overall Activity, it was not included in the Overall Performance measurement.
Data Sources
For this report, we used several sources of data provided by the USPTO:
- USPTO Bulk Patent Data – A set of XML files provided by the patent office, which covers metadata of patents as well as claims, body, and description.
- USPTO Examiner Dataset – The patent office has started releasing examiner data on a yearly basis. This data covers all transactions related to each patent application, the status of applications, applicant and correspondence addresses, etc., often provided in a CSV format or JSON.
- USPTO Examiners Database – A list of all current examiners and their art units as well as technology centers.