Data on Patent Law: Sources and Uses Explained

By Adam J. Feldman • February 1, 2024

Two professionals look at a tablet in a hallway; man points at screen, woman smiles, another person walks behind them.

Sometimes the most useful litigation tools are ones you assemble on your own – that way you can tailor them to your needs, and occasionally they are even free.  Here is an example of resources for the federal circuit, PTAB, and other trial level patent litigation. These resources can give you a sense of judicial behavior which can help generate expectations for case outcomes and timelines.  The three resources that I will quickly run through in this post are the the Compendium of Federal Circuit Decisions compiled by the University of Iowa Law School (what I will call the “Iowa Database”), the USPTO’s datasets and case resources, and CourtListener’s RECAP Archive. Each of these resources is free, and each can significantly assist you in developing a patent rights strategy.

Federal Circuit Database

This Iowa Database is comprehensive of Federal Circuit decisions since 2004 and has multiple pieces of information for each case.  The Database contains 19,761 cases and is consistently updated.  The types of information that one can derive from this dataset are invaluable. Anything from the likelihood of a granted en-banc (136 granted and 13,237 denied for a rate of approximately 1%) to the number of appeals adjudicated from PTAB (1,777) is readily available.  

Since the Iowa Database contains information on all decisions from the Federal Circuit, some sorting is required to isolate particular types of appeals like those relating to patents.  If you have a software application that can easily create crosstabs like Tableau(my favorite) you can organize and synthesize the information to derive useful outputs. The 19,761 records, for instance, can be sorted by dispute type. Although many of the records records relate to orders that aren’t connected to several of the case outcome variables, among cases that are labeled by type, 3,270 deal with patent infringements, 1,059 deal with inter partes review, 517 deal with contract claims, etc.

The patent cases are coded for whether they relate to code sections 102 or 103 along with other issues like claim construction and definiteness.  Once a specific area is nailed down, let’s say patent infringement, then more specific analyses can be performed.  If we isolate the cases from 2015 forward for example, we can see which judges have been the most frequent majority authors (Stoll with 99, then Prost with 93, and Lourie with 87). We can also look to see who authored dissents most frequently (Newman with 31 and Reyna with12). Or perhaps we want to know about the most frequent lower courts (District Court for the District of Delaware with 246, District Court for the Eastern District of Texas with 163, and the District Court for the Northern District of California with 158). Maybe we even want to know who is or was most likely to dissent from an opinion authored by Judge Stoll. In such instances, Judges Dyk, Hughes, Lurie, and Newman each dissented twice.

USPTO Website

The USPTO also has a treasure trove of free resources for the legal data enthusiast. Some of the information is quite helpful for legal practitioners moving forward while other data are mostly historic. Even the backwards looking data though can aid with current decisions to the extent that they are based on litigation before active judges.

The historic element is quite fascinating. While unfortunately only updated through 2016, the Patent Litigation Docket Reports have case level information from 81,350 district court cases filed between 1963 and 2016.  A few nice feature of the Docket Reports is that they track litigation timing and this can be parsed by on other variables like the judge or court of interest.  There are also multiple datasets that correlate to one another so you can look at observations based on the attorneys on the cases, patents, case names, or documents. 

We might, for instance, be interested in the magistrate judges who these cases were referred to in order to gauge how long proceedings end up taking in their courts. Here is an output of magistrate judges with over 200 proceedings in this dataset.

Judge Roy S. Payne for the Eastern District of Texas has the lion’s share of these cases with all other judges only deciding a fraction of Judge Payne’s count. Let’s say we are interested in the time it takes these judges to move from an opened to a closed case, we can use the time parameters in the dataset to run this calculation for each individual case, and then generate averages by judge.  Here are what the averages look like for these judges.

Judge Payne cleared his cases the quickest of the group at just under 250 days while, at the other end of the spectrum, Judge Trumbull of the District Court for the Northern District of California averaged over 535 days per case.

There are also other datasets available on the USPTO site as well including the Patent Examination Research Dataset (PatEx) which covers “13 million publicly-viewable provisional and non-provisional patent applications to the USPTO and over 1 million Patent Cooperation Treaty (PCT) applications.” 

CourtListener’s RECAP Archive

The RECAP Archive is a freely accessible tool that compiles PACER records.  It is an extremely useful resource and was used to derive some of the datapoints for the USPTO measures.

RECAP is generally more of a qualitative data source that can be used to put together quantitative statistics. One of the nice parts of RECAP though is that you can dive into case dockets and in some instances you can view documents filed in cases. 

One of the nice features of the RECAP archive is that you can filter by PACER codes, so, if for instance you were interested in patent cases, you could plug in nature of suit code 830 and find that since the beginning of 2015 there are 28,257 cases that fit under this code and 1,867,132 docket entries. If you were interested in the cases referred to Judge Roy S. Payne in the Eastern District of Texas you could refine your search by judge and find there are 2,611 relevant cases since the beginning of 2015.

A nice feature of RECAP that was presumably used in the creation of the USPTO dataset is the RECAP metadata that correlates with the variables in the USPTO site. These variables include the judge assigned to and referred to the case, the citation, date filed and terminated, date of last known filing, cause of action and nature of suit, jury demand, and jurisdiction type. There are also data on the parties and attorneys where available through PACER.

The upside to these data is that they allow for updating beyond the numbers currently available from the USPTO dataset which only run through 2016 and provide additional information not provided in the dataset. The downside though is that it takes either scraping and parsing skills to put it into a useable format or taking the time to input the data manually. If you have specific information you are trying to assemble rather than raw general data though, this is a good place to begin.

Concluding Thoughts

Legal data help with generating predictions, following trends, and understanding changes in the legal landscape.  The data described in this article are all readily available and relatively easy to use and navigate. These are great starting points for research and comparisons and provide context to those interested in specific cases. Another big upside is that these resources are free.

While the resources I described generally relate to patent law, this is just an example of the legal data that are freely available on the web. There are many other resources for other areas. If you already understand the value of data, then the raw data available to put together novel datasets abound. Furthermore, there are experts in legal data analysis that can help you develop the skills to make use of these resources and to ascertain answers and solutions to complex legal questions that are not answerable through doctrine alone. For claimholders, litigators, litigation funders, and insurers, such data provide the additional benefit of oftentimes lending themselves to probabilistic determinations that can help individuals forecast potential outcomes and generate likelihood intervals that relate to the probability that certain outcomes will come to fruition.

Adam Feldman  is the editor of  Empirical SCOTUS, a blog that conducts data analysis of the United States Supreme Court, and the Principal of Optimized Legal, a legal data/statistical consultancy. He is also an adjunct professor of political science and public law at California State University, Northridge. You can reach  Adam  for specific data and analyses related to your own litigation questions in this and other areas.

< Older Post

Newer Post >

Certum Group Can Help

Get in touch to start discussing options.

Subscribe to Our Newsletter

Newsletter

Recent Content

Chris Seidl Named to IAM Strategy 300 List of Leading IP Strategists

By Certum Group Team • August 13, 2026
Certum Group is pleased to announce that Chris Seidl has been named to the 2026 IAM Strategy 300: The World’s Leading IP Strategists list. IAM Strategy 300 is a global ranking of IP strategists who are leaders in developing and implementing strategies to maximize the value of IP portfolios. IAM identifies individuals through extensive research annually. Chris leads Certum’s IP finance strategy, including IP licensing, litigation funding, and acquisitions. This is the fifth consecutive year Chris has been included on the IAM Strategy 300 list.  Click here to see the complete rankings.

How Lawyers Can Use AI Effectively and Responsibly

By Certum Group Team • August 4, 2026
Artificial intelligence is quickly changing how legal work is researched, drafted, reviewed, and delivered. But while AI can improve efficiency, it also creates serious risks, including inaccurate analysis, fabricated citations, and potential court sanctions.  In this webinar, Certum Group brings together legal and business experts to discuss how lawyers can use AI to strengthen their work without compromising accuracy, professional judgment, or accountability.

Tips for Lawyers’ Effective and Responsible Use of AI

By Certum Group Team • July 29, 2026
Artificial intelligence is transforming all corners of the economy, and the legal profession is no different. At first it seemed the stories about AI and the law were all negative, as many lawyers , even those at some of the most prestigious firms , found themselves sanctioned for filing briefs with hallucinated cases. More recently, the news cycle has turned, as existing law firms embrace AI and new firms sprout to deliver AI-first legal services. Just recently, top lawyers from Kirkland & Ellis and Quinn Emanuel each left to launch their own law firms. Meanwhile, Kirkland, the AmLaw 1 firm, announced it would invest $500 million to develop its own proprietary AI system. [ Click here to read a Bloomberg article by Certum’s Will Marra on what Kirkland’s AI announcement means for the future of third-party legal finance.] This can be dizzying for many lawyers. Clients want them to use artificial intelligence. Competitor law firms are using AI. But the risks of misusing AI are high and can even include sanctions and media coverage that gives lie to the old adage that “all news is good news.” A Transformative New Tool To help lawyers navigate this landscape, Certum Group recently hosted a webinar to help lawyers navigate the landscape. Our featured speaker was Michael Showalter , founder of Showalter PLLC, a litigation firm built around AI tools, and a former appellate lawyer at Gibson Dunn and Wiley Rein. The conversation was led by Suneal Bedi , Certum’s Scholar in Residence who is a professor at Indiana University’s Kelley School of Business. This was the first in a series of conversations that Professor Bedi will lead designed to help Certum’s clients navigate the toughest challenges they face today. In the webinar, Showalter demonstrated the power of AI by sharing several moments that “blew his mind” over the past year: A first draft of a law review article that was better than what he’d get from most junior lawyers. T urning a project that once took 100–150 hours into roughly only 15. A flawless table of authorities, table of contents, and cite-check on a brief, produced in about ten minutes. He noted he’d never once received a flawless table of authorities from a human paralegal. Sophisticated legal reasoning; identifying an overlooked antecedent argument in a forthcoming Yale Law Journal article showing the tools do real analysis, not just “automatable” formulaic work. He now estimates he accomplishes in a single day what would have taken him 40 hours of work back in 2022. Three key takeaways emerged from the webinar. To Be a Good Lawyer With AI, First Be a Good Lawyer AI cannot replace legal judgment, but it can amplify it. Lawyers should not outsource the job of lawyers to an AI model. They should treat the models instead like highly capable but error-prone junior associates. This means they should ensure the model has sufficient context about the legal issue they’re asking it to address. And you should give senior-lawyer leadership and oversight to its work. You should also treat AI errors as inevitable, the same way junior lawyers will inevitably make mistakes. Be vigilant to when the model gets something wrong, fix the error, and do what you can to avoid the error from recurring in the future. And you should iterate constantly. AI workflows cannot be completed in a single prompt. Lawyers should consistently redline and comment on the AI’s outputs to refine and perfect the work product. Guard Against Hallucinations and Errors Lawyers are rightly concerned about the prospect that AI may rely on fabricated cases. The webinar offered some candid commentary on this issue: First, the technology has changed dramatically between 2024 and 2026. The incidence of hallucinated and fabricated cases is much rarer today than in the past, partly because the frontier labs have focused on addressing this problem. Second, verification tools now exist. Tools like Veritas now exist to compare every quote in a brief against its primary source, catching errors rather than predicting text. Third, senior oversight remains non-negotiable. Lawyers should not assume an AI will accurately describe a case any more than they should assume a junior associate will do so. Lawyers should read and review every case they cite for accuracy and reliability. Meet Your Clients’ Expectations and the Courts’ Requirements Finally, it is clear that the market is shifting towards the expectation that lawyers will use AI. Even the most sophisticated clients are now demanding that their clients use AI. And they are relying on the existence of AI to push down rates and demand that simpler tasks get outsourced to artificial intelligence. For example, Sebastian Niles, the President and Chief Legal Officer of Salesforce, recently published an article arguing that the integration of AI into law firms should be a baseline expectation. [Harvey AI’s CEO, Winston Weinberg, was recently the keynote speaker at an NYU Law School conference co-organized by Professor Bedi and Certum’s Will Marra. Click here to learn more about the takeaways from that conference.] At the same time, courts have stepped in to closely police how law firms are using AI. Some courts are even amending local rules to regulate the use of AI in legal filings. To be a great lawyer today and tomorrow, lawyers need to stay on the cutting edge of artificial intelligence. Clients demand it, and courts do too. Lawyers should continue to educate themselves about how to best use AI to improve and amplify their work, but not to replace it.