Predicting Patent Issuance

By Dennis Crouch

One of my patent law students (Marriam Lin) is working on a project that divides patent applications into two sets: (1) Those that are patented and (2) those that are abandoned without being patented. We’re looking to see whether we can identify objective factors found in the original application filing (or at least by the time of publication) correlate with the application being patented or not. This project falls in line with the papers on Valuable Patents and Worthless Patents, but focuses only on pre-issuance issues. We have a few simple measures such as the assignee, type of technology, number of inventors, size of application and claim set, and priority claims. Obviously, our result will have only a limited value since many important factors will elude our measurement. However, I have been surprised at the explanatory power of the model even in our limited approach. If you picked up a patent application, what factors would you think might help predict whether that application will eventually issue as a patent? Is there a way to objectively measure those factors?

31 thoughts on “Predicting Patent Issuance

  1. 31

    Correlate with any second set of eyes actions.

    Oh wait, that star chamber type of information is not available…

  2. 30

    I agree that claim length is a strong predictor.

    If claim 1 is a page and a half long, expect a first action allowance.

    But please look to see if there is a correlation with Examiner Experience.

    I would expect that the more experienced the Examiner, the more likely a patent will issue.

    I believe that inexperienced Examiners are under pressure to reject as rejections imply that they are doing there job and it is the easy position for a supervisor to take with allowances requiring more of a justification from the inexperienced Examiner and with the inexperienced Examiner being unable to articulate a reason for allowance.

  3. 28

    – Word count and number of elements in independent claims, by class code
    – Inventor location (multi-country inventors may correspond to global tech companies, as opposed to US domestic firms)
    – whether an assignment was filed or not
    – Patentee as lexicographer – use of non-standard terms within claims (e.g. “a vector graphics processing unit” vs. “a graphics processor”) and # of words that make up nouns in the independent claims (e.g. a claim with “processor”,”memory”, and “cache controller” may be less likely to issue than “a multi-core special purpose processor”, “a direct-addressable memory”).

  4. 27

    Part of the Capos regime frauds against inventors is the lack of predictability of patents filed. Instead of treating the intellectual property as owned by the filer as it should be It becomes anyones floating ip with an indegent inventor loosing everything when unable to attract investors

  5. 26

    Type(s)/raw number of FAOM rejection(s): 101, 102, 103, 112.

    Whether same/different rejection(s) in subsequent rejection(s).

    Whether con(s) and/or div(s) relying on subject app are filed subsequent to FAOM.

  6. 25

    First you can note the degrees of specificity used in the terms of the claims. You can note the length of the claims. You can combine the aforementioned.

    Pencil test – see above.

    Also, this is covered by the “How to get a picture claim that is worthless” technique.

  7. 24

    combined with their general tendency to be weaker at math than English

    It’s opposite day.

  8. 23

    Oh and there is another way to tell at the beginning of prosecution whether or not a patent will eventually issue. Look to see if there is a 102b for the entire application in an IDS. Lol

  9. 22

    “f you picked up a patent application, what factors would you think might help predict whether that application will eventually issue as a patent? Is there a way to objectively measure those factors? ”

    Yes there are ways. In my art at least. Everything I’m about to mention goes for the initial claims and is discussing allowability on first action, but can equally be applied to the subject matter within the spec as well. And of course if the application is allowed on first action then the prediction of whether you get a first action allowance is a very good prediction of whether the application will eventually issue as a patent.

    First you can note the degrees of specificity used in the terms of the claims. You can note the length of the claims. You can combine the aforementioned.

    You could theoretically also compare the prevalence of very specific and distinctly non-generic words used in the claim with how often they are used in the art. (that is, measure the degree of “exoticness”, or the ubiquitousness, of the subject matter)

    You can note the size of the subclass into which they are classified and how well the case fits in with the definition of the subclass that they are properly classified under.

    You can note the degree of complexity or simplicity of the subject matter (on a rough scale). Anything extremely complex is more likely to be allowable. And indeed, sometimes something of extreme simplicity can also be more allowable (because of how easy it should relatively be to more complex things).

    You can note whether or not the case goes “up and down” the “ladder” of design, for instance, specifying some very small components as well as specifying a larger machine made up of those components. This is especially predictive when the subject matter of the small parts is classified differently than the large machine.

    You can oftentimes tell from the initial claims how subjectively willing the applicant is to accept more narrow claims based on objective indicia. That is, look at the number and variety of independent claims of widely differing scope all directed to the same species and group (taking away the possibility of restriction between the inds).

    You can also tell how subjectively likely the applicant is to accept narrower claims by where the case is classified, both in terms of classes and subclasses. In my art at least there are some subclasses with 5000+ references all about a very specific little part of a machine. Anyone trying to get a patent there is much more likely to be receptive to narrow coverage of their invention compared to someone operating in a subclass where there are less than 100 references and all of those are completely different from each other. On the other hand, looking at class classification instead of subclasses within my art, I have noticed that when I do cases of relatively simple arts the applicant’s representatives will be notably less pushy towards trying to obtain optimum coverage and will be more concerned with just getting a patent and so of course they can usually get some patent, though it will be narrow in scope.

    The list goes on and on, I could just write these for awhile.

    Indeed, these are some objective factors that are considered in the actual allowing of a case. When you are a master of all of these objective factors it becomes much easier to get cases by a primary or SPE just fyi for new examiners.

  10. 20

    It’s certainly true that correlation does not equal causation. But it’s equally true that correlations may provide clues to causation.

    And given lawyers’ obsessions with “risk” combined with their general tendency to be weaker at math than English, if there was evidence that applications filed on an odd day were granted with, say, twice the likelihood of applications filed on even days, you can rest assured there’d be a few lawyers advising their clients accordingly.

  11. 19

    There are more patent granted an odd day than an even day.

    The chances that your patent is granted an odd day are greater than an even day.

    Does the fact that your patent is filed an odd day increase your chances that the patent be granted?

  12. 18

    I agree that there would be a positive correlation with PCT filing. And by extension there should be an even stronger positive correlation with national/regional filings in foreign countries.

  13. 16

    Obviously the most important issue leading to patent issue should be marketability of the subject matter and is it practicle or impractical and therefor worthless. I here Kapos regimes calls for patent filing perfection or rejection and the allowance of theft by other filers that destroys the original concievers incentive to create subsiquently shuttering the patent system of human advancement very disgusting corruption!

  14. 15

    If you picked up a patent application, what factors would you think might help predict whether that application will eventually issue as a patent? Is there a way to objectively measure those factors?


    If you find a correlation between length and allowance, we will all be obliged to stuff 200 paragraphs and 50 sheets of drawings into every application, which in my experience correlates positively to annoying the examiner and negatively to allowance.

  15. 13

    Age of inventor – Some of mine have been pending for over 10 years, so an inventor could very well die before anything ever gets allowed.

    Wealth of inventor – As y’all know, it takes a bunch of money to get a patent allowed. I suspect that most folks realize after the first decade of waiting at the PTO that it really is a futile effort and, if they read this blog, the cases show that it is not very likely to return anything anyway… might as well go fishing.

  16. 12

    Given human nature, the applications most likely to be abandoned are those with the least sunk costs and those claiming priority to post-GATT applications nearing their expiration date (i.e., those with the shortest time remaining for recouping investement). So I would guess the most predictive criteria would be the length of the application (reflecting the cost) and the date of filing relative to the priority date (later filing dates -> more likely abandonment).

    The number of granted patents claiming the same priority application may correlate with the latter of these measures and may therefore also be predictive. On the other hand, if a patentee is maintaining those patents, the cost of keeping an application pending is relatively small and the “flexibility” afforded by the pending application may be enough to justify its pendency even if there is very little remaining in the disclosure to be protected.

  17. 11


    Total claim count — as originally filed.

    Total claim count — not as originally filed.

    Same two questions; but with independent … and dependent claim counts.

    Existence of/number of amendments and/or RCEs.

    Existence/non existence of one or more priority filings, e.g. provisionals, PCT, etc.

  18. 10

    Quality of the patent application as filed — which also indicates that it was screened by a competent patent attorney.

  19. 9

    Many pro se applications contain the text string “Conclusion, Ramifications, and Scope” as recommended in David Pressman’s Patent it Yourself book. I found it helpful in trying to dig out examples of poorly written provisional applications.

  20. 8

    Single inventor with no assignee and claiming SES. It wouldn’t work for this study, but in the future a single inventor with no assignee and claiming micro-entity status would be even more determinative.

  21. 7

    Try examining a correlation between (# pages in spec/# of figures) vs allowances. I would think the higher density of figures, the more features that could possibly be found patentable.

  22. 6

    Independent inventors…yes…but how would Dennis test for that? Single inventor with same name as applicant?

    Small entity status might also show some negative correlation.

  23. 4

    Positive correlations for issuance: university as assignee; PCT filing; priority claim to an earlier application that has since issued. That last one has the caveat that the effect is probably weaker—and may even skew negative—for later applications in large patent families as a) the outermost bounds of the scope of the original application are worked out and b) the applicant figures out what claims the market actually cares about and may abandon the rest.

    Negative correlations: independent inventors, particularly pro se; “bad” claim structure (e.g., multiple dependent claims).

    Any chance you’ll try to take into account the attorney of record? Some firms screen clients more than others, and some take more care than others in handling applications. I think that would show up in issuance rates.

    I assume paying for any of the types of accelerated examination is public knowledge, though I’ve never actually looked. I expect that for the relatively small number of applications in that category they would tend to issue at a higher rate than average.

  24. 3

    Independent claim word count and word count in general of the initially filed claim set (being indicative of narrower initial scope).

  25. 2

    Whether an IDS is submitted with the initial filing and number of references in the IDS

  26. 1

    In my experience the most reliable predictor of issuance is the amount of detail in the specification, particularly, the number of disclosed embodiments.

Comments are closed.