Well, this is interesting (and quite surprising). A new study by John Allison, Mark Lemley, Joshua Walker looked at highly litigated patents to see how well they did in court
, and came up with some really unexpected findings. The summary of the findings are as follows:
We designed this study to explore the effects of repeat play on litigation behavior, contributing to a literature on the economics of civil procedure as well as the substance of patent law. But what we found was dramatic and unexpected: The patents and patentees that occupy the most time and attention in court and in public policy debates -- the very patents that economists consider the most valuable -- are astonishingly weak. Non-practicing entities and software patentees almost never win their cases. That may be a good thing, if you believe that most software patents are bad or that NPEs are bad for society. But it certainly means that the patent system is wasting more of its time than expected dealing with weak patents. And it also suggests that both our measures of patent value and our theories of litigation behavior need some serious reconsideration.
Digging deeper into the report, it looks at and tests a variety of different concepts around patents and litigation. In theory, if a patent is used in multiple patent cases, you tend to think that it must be a pretty solid patent, and one that has been vetted plenty of times. And yet, when the researchers looked at the 106 patents that have been involved in eight or more lawsuits since 2000, they found that the patent holder wins such cases only 10.7% of the time. For patents that have only been brought to litigation once, the patent holder wins 47.3% of the time -- an astounding difference.
My first thought on hearing such numbers is that the data could be misleading in that many companies may be a lot more willing to settle when sued by a serial patent litigator. However, the researchers tested that and while they did find that a higher percentage of those sued will settle in cases involving a "most-litigated" patent as compared to a "once-litigated" patent, the higher settlement rates don't offset the huge difference in win rates.
The study also looked at the differences in entities doing the suing as well as the types of patents involved in the lawsuits. And, once again, there are some interesting finding. Patents that initially came from big companies "represent only 22.4% of the assertions in the most-litigated group." I don't find this all that surprising. Big companies tend not to be serial
patent litigators over the same patent. There certainly are some exceptions, but in general, it does seem like big companies tend to view proactively suing for patent infringement as a "last resort." However, patent hoarders and non-practicing entities who base their entire business model on demanding money from patents are probably a lot more likely to sue. But where it gets interesting, is the fact that they're also a lot more likely to lose
their patent infringement lawsuits: only 12.3% of "small entity plaintiffs" win their lawsuits (excluding default judgments).
Of course, size doesn't necessarily indicate whether or not the company is a "product" company or a non-practicing entity (commonly called a patent troll). In the "most-litigated" patent group, not surprisingly only 16.7% of the assertions came from actual product companies -- meaning a whopping 83.3% of the assertions in the most-litigated group came from patent trolls. So how do the trolls do? Not so well. If they fail to settle, patent trolls win just 8% of their lawsuits (9.2% if you include default judgments). For comparison's sake, companies that actually make stuff, win 40% of their cases and 50% if you include defaults. It certainly sounds like juries are a lot more sympathetic to companies that actually make stuff.
Of course, it's important to point out that patent lawsuits almost always settle. The settlement rates tend to hover near 90%, because the cost of litigation is often so high that it's cheaper to settle. But, if anything, this data suggests that patent trolls probably are better off not pursuing the lawsuit if they can't get the companies they sue to settle (and, correspondingly, it may make more sense for those sued to actually fight back).
The final part of the study breaks down the different patents by type -- and provides some interesting info for those who dislike software patents. Once again, litigating frequently over a software patent tends to be a loser's game -- even though software patents were 93.7% of the most-litigated assertions. And yet... software patent litigants in the most-litigated group only win 12.9% (with or without default judgments) of the time. Non-software patents, however, win at much higher rates: 37.1% without defaults and 51.10% if you include defaults. In other words, the feeling most people have is probably true: the software patents that we see in various lawsuits tend to be of dreadful quality
, and the results of the lawsuits support that view.
The authors do try to analyze the data -- and one key missing data set in all this are licensing
deals outside of settlements. Oftentimes, lawsuits won't even be filed if companies realize it's cheaper to pay up on the threat
of litigation. From there, the authors note that once a patent has a few licensing agreements, the patent holder will probably be a lot less interested in suing anyone else, because if the patent gets invalidated, that patent holder will likely lose the licensing revenue stream.
As for what this means for the overall patent system, the authors note two possible (contradicting) interpretations, and then split the difference:
On the one hand, it should give substantial ammunition to those who argue against software patents and who want to restrain patent trolls. If software and NPE patents are overwhelmingly bad -- either invalid or overclaimed -- the social benefit of allowing them may well be outweighed by the harm they cause. At the same time, however, one could read this evidence as proof that the system is working -- that the bad patents are being weeded out of the system and are not stifling innovation.
The truth probably lies somewhere in between. The latter claim -- that the widespread invalidation of software and NPE-owned patents shows that the system is working -- seems altogether too facile. After all, roughly 90% of those cases settled without judgment. While those settlements are confidential, we expect that the vast majority involved some sort of payment to the patent plaintiff -- a payment that the outcomes data suggests might represent not the acquisition of real legal rights but a nuisance settlement over a likely-invalid patent. At the same time, the fact that these patents are so weak should -- at least once exposed -- limit the value of those settlements, and quiet concerns that software or troll patentees will actually shut down very many innovative products.
I'd argue that when you include license agreements based on threats of legal action along with all the settlements it helps to dilute the value of the data even more, to suggest the idea that the "system is working" is false. The number of companies who pay up without a legal ruling is many, many, many times the number who actually go through to the end of a lawsuit. And, while the authors hope that exposing this data might convince more companies to push back against weak patents, I'm not convinced that will actually happen. It's still going to be cheaper for many companies to just take out a license or to settle than to fight. And, finally, as the authors point out later in the report, when you see the massive size of a few patent awards, the patent holders may view them as lottery tickets. Quoting the paper:
If the payoff for victory at the end of the day is $1 billion, a 10% chance of success doesn't sound so bad.
On the whole, the results certainly seem to suggest that patent trolls with software patents do very much view the system as a lottery ticket, and they're willing to use really weak patents to try to win that prize. That is not at all what the patent system is designed to do, but it's how the incentives have been structured -- and that seems like a pretty big problem that isn't solved just by showing how many of these lawsuits fail. The amount of time and resources wasted on those lawsuits, as well as the number of companies who pay up without completing a lawsuit, suggest that there is still a major problem to be dealt with.