Who appeals (and wins) patent cases?

By Jason Rantanen

There are lots of studies of Federal Circuit decisions, but very few involve the link between all cases filed at the district court cases and appeals. This prompted the question for me: who actually files appeals in patent infringement cases and how representative are they of the underlying civil actions filed in the courts? It turns out that the answer is “mostly patent asserters” and that they aren’t necessarily representative of case filings.

One of my research team’s ongoing projects is the Federal Circuit Dataset Project, which involves collecting all of the opinions & orders that the Federal Circuit makes available on its website, coding information about them, and then putting it all together into a publicly-accessible database that anyone can use to answer their own questions about the Federal Circuit.  There’s a user-friendly portal to the decisions dataset at empirical.law.uiowa.edu, and the full dataset is available at https://dataverse.harvard.edu/dataverse/CAFC_Dataset_Project.

This year, our major project involved linking up the CAFC dataset with datasets of patent infringement cases filed in the district courts. A draft of the first paper to spin out of this project is now up on SSRN at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4162979. (There is an option to download without logging in, but sometimes you have to hunt for it on the page.)  Thanks to the Houston Law Center’s Institute for Intellectual Property and Information Law for inviting me to present it at their national conference in June. Some takeaways:

  • For patent infringement cases filed between 2011 and 2016, about 6% have at least one appeal, although this rate has been declining from 7.8% of cases filed in 2011 to 4.4% of cases filed in 2016.
  • The average time from case filing to first appeal is 27 months, although there is substantial variation (SD of 18 months with rightward skew).
  • Appeals are overwhelmingly filed by patent asserters: 82% of first appeals in the set were filed by patent asserters compared to 18% by accused infringers.

For this particular project, we linked the Federal Circuit dataset to the Stanford NPE Litigation Database, which contains coding on patent asserter types. (Details about the Stanford dataset are available in Miller et al., Who’s Suing Us? Decoding Patent Plaintiffs since 2000 with the Stanford NPE Litigation Dataset, 21 Stan. Tech. L. Rev. 235 (2018)). Using this data, we found some differences between the frequencies of case filings and appeals.  Figure 2 from the paper shows the most frequent cases by Stanford litigation dataset category. Category 1 are companies that acquired patents but do not themselves make any products or provide any services.

Table of case filer frequency

Figure 2 from Who Appeals Patent Cases

Figure 2 shows that while companies that acquired patents (Stanford litigation dataset category 1) filed 37% of patent infringement actions from 2011-2016, only 25% of appeals from cases filed between 2011 and 2016 arose from those cases. On the other hand, while product companies accounted for only slightly more patent infringement lawsuits (just over 40%), appeals in cases filed by product companies constituted 54% of all appeals filed. In other words, cases brought by product companies are appealed at a higher rater than cases brought by companies that acquired patents. This is consistent with prior findings that cases brought by PAEs settle at a higher rate than cases brought by product companies. Not only that, but we also found a difference in who is bringing the appeal in these cases: in cases brought by a PAE, the patent asserter was the primary appellant about 90% of the time, whereas in cases brought by product companies the patent asserter was the primary appellant only about 70% of the time.

What about who wins these cases?  That’s a bit more complex, and depends on how you define “win.” We define a “win” as obtaining an affirmance if a party is the appellee and an affirmance-in-part, reversal or vacate if a party is the appellant.  Using this definition, we found that overall, patent asserters usually lose on appeal, with product companies being much more successful than PAEs. However, this is heavily affected by the fact that both patent asserters generally and PAEs specifically are in the shoes of the appellant most of the time, and more often than not, the Federal Circuit affirms the district court.

The paper itself goes into more detail, including information about affirmance rates by appellant type, other categories, and the kinds of forms that these decisions take.  You can read it here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4162979.

Disclosure: I am receiving an honorarium from the University of Houston Law Center Institute for Intellectual Property & Information Law upon publication of this paper in the Houston Law Review.

43 thoughts on “Who appeals (and wins) patent cases?

  1. 7

    Seems the system is a bit unbalanced.

  2. 6

    Off topic, but CPC Patent Technologies PTY Ltd has petitioned the Court for review of Apple’s mandamus transfer out of the WD Tex.

    Read more at: link to law360.com

  3. 5

    This article author’s blog comment: “That’s exactly right – the [6% appeals] denominator is [from] all cases filed, and most cases settle before a decision. The next step is looking at actual decisions at the district court – the appeal rate for those is much, much higher.”
    This is another confirmation of how few patent suits ever even get to trial these days, and that those that do usually get appealed. Few [4-5%?] get pre-trial summary judgements based on prior art, and many of those get remanded on appeal for disputed facts, as prior statistics from the Houston Law Center had shown. As distinguished these days from more frequent S.J.s for non-infringement (per Markmans) or Alice-type unpatentable subject matter. I.e., most, but not all, final judgments below are settlements.]

  4. 4

    To get a bigger sense of the full patent landscape, you would have to have the data on how many patents are licensed without resort to litigation at all. Nobody has a full view of this data set, which is confidential and unknowable. As “The Prophet” has indicated, litigation happens (not always, but often) when the defendant doesn’t want to take a license because it thinks the patent is invalid, not infringed, or both. Thus, the contested cases are always going to favor the defendant–that is by design. People that either blindly think “all patents are good” or “all patents are bad” are wrong.

    1. 4.1

      Wrong. People that blindly think that “all patents are good” or “all patents are bad” are both correct.

      1. 4.1.1


      2. 4.1.2

        You have to realize D416,453 is a SPATULA. So D416,453 is invalid by DP. So moving my IP to the Assignee NESCCO makes the move MOOT.


          I wonder though was Guerwood told that with the switch would the NDA protect my maternal family and Guerwood’s family too. If you ask me that NDA was the NDA that covered lots and lots of situations. Richard Egan assumes I’m short of a few cards, I am sure he is. Imagine him telling me in His Jackson kitchen, never realizing I would put it together.

    2. 4.2

      Fair point.

    3. 4.3

      when the defendant doesn’t want to take a license because it thinks the patent is invalid, not infringed, or both.

      You missed the largest driver – naturally.

      1. 4.3.1

        I think the proper way to articulate it is:

        When the infringer doesn’t to take a license because they think that patents in the USA can rarely be enforced no matter how high the quality the patent is or how great of an invention is claimed. Moreover, even if the patent is enforced it will be many years before a final judgement is reached and the final judgement will likely be about what we would have had to play in a license anyway. And there ain’t no stinking injunctions no more after ebay.


          That’s pretty good paraphrase of the Rational Actor view of Efficient Infringement.

          Throw lots in the way of the patent holder – forcing them through post grant review and chasing you through the courts, and if at the next of the day, the patent holder holds on and wins, then there is still the (good) chance that all you have to pay would have been what you should have paid to begin with.

          Add in that a single loss for the patent holder is devastation, and what Rational Actor (with no respect for the Rule of Law) would act otherwise?


          it will be many years before a final judgement is reached and the final judgement will likely be about what we would have had to play in a license anyway.

          Except that under the American rule, the infringer will have a giant pile of lawyer bills to pay, along with that judgment that roughly approximates what s/he would have paid if s/he had just negotiated the license beforehand.

          there ain’t no stinking injunctions no more after ebay.

          This just is not true. “eBay has effectively created a bifurcated regime for patent remedies, as operating companies who compete against an infringer still obtain permanent injunctions in the vast majority of cases that are successfully litigated to judgment. In contrast, non-competitors and other non-practicing entities are generally denied injunctive relief.”


            I’d like to see the numbers on that. Never seen any stats to back that up.


            OK Greg the attorney fees are there.

            So that plays into the calculous of whether to take a license or not. But as anon points out there are many ways to invalidation IPR, and DC, CAFC.

            So $5 million in attorney fees but they want a license for $50 million. That $5 million isn’t going to change the decision.

            The attorney fees only come into play when the license fee is lower than like $20 million. And can swamp the entire calculous when the license is only like $1 million. Still, the patentee is likely going to get an IPR where 80 percent of petitions are granted and 70 percent of all petitions result in at least one claim being invalidated.


              [In] IPR… 80 percent of petitions are granted…

              Not even close. IP Watchdog used LexMachina a few years back to track a cohort of petitions through to the end. Only 50% of those petitions were instituted.


                link to ipwatchdog.com

                Institutes 70% and 80% of those have at least one claim invalidated.

                1. Did you click through the links in that post. The numbers they cite do not really bear out their claims. First, that “70%” figure is not any sort of historical average. It comes from looking at the “FY22 YTD” column. In any event, those data are stale. The latest figures show a 65% rate, not 70%.

                  Second, that column contains 138 instituted petitions, and 71 denied petitions. In other words, your assertion that “the patentee is likely going to get an IPR where [70] percent of petitions are granted” depends on a sample size of 138. There were 135 new suits filed just in Q2 of this year. In other words, your data depend on a woefully inadequate N on which to base any conclusions.

                2. To be fair, however, I will concede that you said 80%, while I said 50%. The latest numbers say 65%. We were both equally wrong (each off by 15 percentage points). I learned something from following up on this question, so I thank you for the lesson.


              >OK Greg the attorney fees are there.

              Theoretically, anyway. But, in practice, that’s just a matter of spending $25K for an opinion.

              side note: attorney fees are expressly limited to “exceptional cases.” Unlike enhanced damages.


                You are talking about an award of the other side’s fees. I am just talking about the defendant’s own fees. The defendant’s lawyers do not work for free, or even all that inexpensively.

                If a company seriously adopts a policy of “I will never take a license unless and until I am legally compelled to do so by an adverse judgment,” that company is setting itself up to waste a lot of good money paying lawyers to fight losing battles. Not surprisingly, very few companies take such a policy—notwithstanding the fevered delusions of the disgruntled of these boards.


              $5 million in attorney fees but they want a license for $50 million. That $5 million isn’t going to change the decision.

              Two point:

              (1) $5 million here, $5 million there, pretty soon you are talking about a lot of money. Meanwhile, the $50 million that you pay for the license buys value. The $5 million + $5 million (etc) that one pays to fight a losing battle is pure waste. That is why few enterprises have a policy of refusing to license in the face of a credible patent infringement charge.

              (2) It is a mistake to imagine that the costs of pointless litigation accrue only in attorneys fees. Being involved in litigation consumes time of the C-level. No company with an actual product to get to market can afford to devote >50% of its executive bandwidth to litigation on a chronic basis, which is what would come of the never-license position that you imagine to be commonplace.


                “Commonplace” is a bit of a red herring.

                That such “no deal with Tr011s” platforms DO exist is without question.

                The “better” question is WHY is Greg so furiously attempting to support one particular side here (and that side being the Efficient Infringers)…?


          Real companies accused of infringement represented by normal patent litigators are not counseled to deny the existence of frequent decisions of up to treble damages and even attorney fee sanctions for willful infringement and/or meritless defense conduct and the continued availability of injunctions and ITC import exclusion proceedings for suing companies inventing and producing real products [as opposed to the large number of suits by PAEs as demonstrated here in this study].


            Please stop the propaganda for the “must make” type of patent holders.

            The patent right is expressly a negative right.

            You continue to lose credibility Paul with your non-stop spin to promote infringement against those that do not make.

    4. 4.4

      Re “data on how many patents are licensed without resort to litigation at all. Nobody has a full view of this data set”
      True, just a strong personal impression that it has significantly decreased in recent generations in favor of “sue first and negotiate later,” which I have seen in almost all of a substantial number of PAE suits I have personally know about and know were settled after suit in in the last 20 years. And I almost neve saw that before that in patents asserted by one major company against another. [Anyone with different actual observations is free to disagree.] One reason, of course, was for the patent owner to insure getting the suit into a highly favored district court – First EDTX and now Waco WDTX.

      1. 4.4.1

        More pandering from you Paul.

        Put the pompoms down.

  5. 3

    Just a friendly reminder to the squirrels below and their nuts: litigation happens when the patent at issue fails to impress the defendant, who chooses not to license the patent. At that point, it’s up to the patentee to decide whether to take the case to court. The patentee’s choice, of course, is colored by all kinds of considerations but one of those considerations is “Do I have anything better to do with my time?”

    The point is that the types of patents and the types of patentees that end up in the courts are already a selected group and the selection in question definitely is not for the “best” patents or the “best” patentees.

    1. 3.1

      While that may be true for some patentees I just don’t think that’s how it works in large industrial settings where each side has dozens or hundreds of patents they’re asserting. In my own art, in the hundred or so cases I’ve seen that went to litigation only like maybe possibly up to 5 were all that suspect to me and all of them were decent little inventions, though perhaps with somewhat broad claims at issue.

    2. 3.2

      How does your “fails to impress” fit in with the Efficient Infringer doctrine?

      Are you blithely unaware that “getting away with infringement” is a major driver of “failing to impress?”

  6. 2

    >For patent infringement cases filed between 2011 and 2016, about 6% have at least one appeal,

    Presumably, that’s 6% of filed cases (i.e., settlements are included in the denominator) …. otherwise, it’s ~15x less than I’d would have predicted.

    1. 2.1

      That’s exactly right – the denominator is all cases filed, and most cases settle before a decision. The next step is looking at actual decisions at the district court – the appeal rate for those is much, much higher.

      1. 2.1.1

        Is there any type of adjustment to account for the AIA’s disparate treatment of non-practicing entities (artificially inflating that denominator)?

  7. 1

    I feel like this is just pretty good evidence that patentees lose most of the time, and thus appeal most of the time. It also refutes (a bit) the notion that the patent system is going strong because look at how many patents there are.

    It would be interesting to see evidence of patentee appeals after a split decision (where the patentee wins some, but not all of the relief it seeks) and compare that to appeals by accused infringers in the same boat.

    1. 1.1

      That is my question as well. Fig 4 in the paper shows that the patent asserter is the initial appellant in 79% of appeals. Is that because the patent asserter loses 79% of cases in the lower tribunal? Or do patent asserts win 50% of the time, but they are much more likely than accused infringers to appeal a loss?

      1. 1.1.1

        I see on pg. 32 that “several studies have shown that patent owners only win the entire case (as opposed to a particular issue) about 25% of the time.” So the 79% of appeals is only slightly disproportionate the the asserters’ ~75% loss rate at the tribunal below.


          Right – so this is confirmatory for the most part, but it means that patentees are slightly more likely to appeal when they lose than accused infringers are (probably because accused infringers can pay and settle if they lose, while patentees are unlikely to receive anything in settlement if they lose)


            Makes sense.

    2. 1.2

      [P]atentees lose most of the time, and thus appeal most of the time.

      Evidently. Prof. R addresses this point around pg. 32 of the draft, where he observes that “a patent owner must win on every issue—infringement and the multifaceted validity requirement—while an accused infringer need only win on one issue in order to prevail.”

      That makes a certain measure of sense. A patentee needs the claim to be valid, and to read on the accused product. An accused infringer can escape liability by losing on either of those points, but winning on the other. The patentee’s mountain is just steeper to climb than the defendant’s.

      1. 1.2.1

        Greg, the patent owner only has the burden of proof, by only a mere preponderance, for infringement. Defendants have a very difficult “clear and convincing evidence” burden to a lay jury on almost everything else. Except, now, for patent or publication prior art challenges in IPRs. That has been a significant change in avoiding trials, and thus appeals, that is not discussed immediately above – is that discussed in the article itself?
        The statistic I found most interesting at first glance is the proof that PAEs suits on acquired patents are now so common that they have nearly reached the level of patent suits by product companies.


          (1) The defendant’s counterargument against infringement only has to be preponderance-of-evidence. It is only on invalidity that the defendant has the clear-and-convincing standard.

          (2) Any more, the defendant often as not gets to argue invalidity under the preponderance-of-evidence standard in IPR/PGR.

          (3) The actual cash value of the difference between preponderance and C&C is fairly small. Most of the time, evidence that will meet one standard will also meet the other. It is the rare and exceptional case that actually turns on the difference between preponderance-of-evidence and clear-&-convincing.


          Talk about spin…

          The statistic I found most interesting at first glance is the proof that PAEs suits on acquired patents are now so common that they have nearly reached the level of patent suits by product companies.

          From your normal “0h N0es, Tr011s” relentless stream of propaganda, I actually had the exact opposite reaction to seeing that the stats showed LESS THAN 50% of suits are from non practicing entities.

          Here (somehow) you want to turn that number into an “alarm.”

          Enough already Paul – your MISinformation is caustic. YOU harm innovation protection with this mantra.

    3. 1.3

      Some complicating percentage of appeals are to try to increase or reduce the damages award base or amount, obtain or avoid enhanced [up to treble] damages, or attorney fee sanctions, rather than getting the infringement decision reversed. Those can be “where the patentee wins some, but not all, of the relief it seeks,” but would you call them “split decisions”?

Leave a Reply

Your email address will not be published. Required fields are marked *

You can click here to Subscribe without commenting

Add a picture