20 July 2010

Statistics II: How Many Examiners Does the USPTO Need?

For this second in a short series of data-heavy reports, we set out to examine the USPTO's performance data in some detail, in an effort to determine how realistic (or otherwise) are the goals and benchmarks set out in the recent Draft USPTO Strategic Plan for FY 2010-2015.  We confess that we did not commence the exercise with high expectations, and we anticipated that we might produce the common type of cynical report that often follows these kinds of crystal-ball-staring exercises.

However, we have ended up forming a view that the USPTO's goals, in terms of reduction of the backlog of pending applications, and achieving and maintaining low pendency time, are quite attainable.  Furthermore, the increase in size of the examining corps proposed in the Strategic Plan is more than adequate to achieve these goals.  A number of current proposals, pilot programs and other initiatives are all likely to have a positive impact.



Over the 2009 calendar year, approximately 456,000 utility patent applications were filed at the USPTO.  Of these, some 240,000 (ie over half) were of foreign origin.  While the total number of applications filed in 2007 and 2008 was almost identical, the number of those that were foreign-originating was 215,000 and 225,000 respectively.

In 1963 the proportion of patents granted to foreign applicants was 18%.  In 2008 this figure hit 50%, and based on current filing rates we can expect this figure to continue to grow.

It is clear that in a globalised economy, the health and efficiency of the US patent system is of concern not only to US residents, but to the whole world.  We demonstrated previously that the now-defunct "Final Rules", seeking to restrict continuations, requests for continued examination (RCE), and overall claim numbers, struck directly at the applicant behaviours having the greatest impact upon the USPTO's ability to reduce the backlog.  However, they would have achieved this at the expense of sacrificing some of the genuine needs of users of the patent system.

Director David Kappos is therefore to be commended for abandoning the Final Rules, and seeking to address the backlog of over 700,000 applications awaiting a first office action in a variety of more creative ways.  His consultative approach to the process is also admirable.  (On this point, we note in passing that the public meeting on the proposed three-track examination process is to be held this Tuesday, 20 July 2010, and will be simultaneously webcast - details here.)

The United States is one of the few jurisdictions in the developed world that has historically provided applicants with very little control over the progress of their patent applications.  Relatively accessible expedited examination procedures are available in a number of patent offices, including those of Europe and Australia.  Additionally, in jurisdictions with a seperate examination request requirement (eg Australia, Canada, Europe, Japan, China, South Korea) applicants may elect to file an early request for examination, rather than waiting for the deadline to do so (which arises between three and seven years from filing, depending on the national laws).

Until the advent of recent initiatives, such as the various Patent Prosecution Highway (PPH) programs, the Green Technologies initiative and Project Exchange, the only available avenues for acceleration of processing were generally considered to be either excessively onerous, or unreasonably prejudicial to applicants' interests, or both.


Ultimately, however, the backlog of applications can only be reduced and then, more importantly, kept at bay, if there are enough examiners to handle the overall workload.  At the end of fiscal year 2009, the USPTO employed 6,143 examiners.  The recently-released Draft USPTO Strategic Plan for FY 2010-2015 proposes hiring 250 "IP-experienced" examiners in FY 2010, and a further 1000 in each of FY 2011 and 2012.  However, with the attrition rate in FY 2009 being around 500 examiners, unless this can be successfully addressed as well, it is possible that the size of the examination corps by the end of FY 2012 may be no more than about 7,500.  This does not seem, on its face, like enough to cope with a 700,000 application backlog.

Simple Queueing Model

We have described a simple model of the USPTO backlog as a queueing system, in which the filing of new applications represents "arrivals" to the queue, and the disposal of applications by the examining corps represents "service" of the queue.  In this respect, both "arrival rate" and "service time" are statistically-distributed, but at any given point in time each has a mean value that depends, in the simplest case, on the rate at which applicants are filing new applications, and the size and efficiency of the examining corps.

In our simple model, we proposed treating RCEs as new arrivals, on the basis that the filing of an RCE effectively restarts the Examiner's clock for processing of the application.  Our own experience suggests, as one might expect, that the Examiner does not have to do as much work after an RCE as prior to a first action on the merits, due to being already familiar with the application.  However, this probably does not alter the general utility of the model, since the result may be simply to bring the effective average service time down a little.

Parameters of the Model

In order to apply the model, we need to know, during some representative period:
  1. the number of new applications filed;
  2. the number RCEs filed; and
  3. either the average service time (ie aggregate time spent by an examiner from picking up an application for the first time, until abandonment or the filing of an RCE by the applicant) or the number of services completed (ie allowances and Final Actions leading to rejections or RCEs).
Looking at the 2009 fiscal year, there were 458,000 new applications filed.  In addition, the number of RCEs filed was around 38% of this number, ie about 174,000.  In our model, therefore, the total number of "arrivals" for FY 2009 was 632,000, or about 12,150 per week.

A more difficult statistic to obtain is the average "service time", ie total time spent by an Examiner on each application.  In their book The Patent Crisis and How the Courts Can Solve It Dan Burk and Mark Lemley use a number of sources to arrive at a likely figure of around 18 hours, although it would appear that the time varies significantly depending upon the technology, perhaps between 9 hours and 32 hours (Burk and Lemley, p 23 n 10). 

Based on these estimates, and assuming that examiners have other responsibilities in the Office, we might assume that the service rate is about two applications per week per examiner.  With 6,143 examiners. this results in a total average service rate of 12,285 applications per week.

Following the alternative "number of services" approach, the total number of utility patent disposals for FY 2009 was around 485,000.  Again adding to this the 174,000 "services" that led to RCEs, the total is 659,000, equivalent to 12,800 applications per week, or 2.06 per examiner per week.

Both approaches therefore give roughly the same result, which gives us some confidence that our assumptions are reasonable, and that we are not missing any significant data.

Both figures are also higher than our estimated arrival rate of 12,150 per week.  Since this "excess capacity" can be used to address the backlog, we would expect to see the total number of applications awaiting examination to be slightly reduced over this period. 

Impact on the Backlog

In the 18 months between January 2009 and June 2010 the reduction in backlog was around 30,000 applications (from 765,000 to 735,000), equivalent to about 20,000 applications per year.  The service rate required in order to achieve this reduction is 12,535 applications per week, which is right around the middle of our two estimates.  All of our data therefore appears to be consistent.

For pendency to a first office action to be 10 months (the proposed benchmark in the USPTO Strategic Plan), the backlog at the end of FY 2009 would need to have been around 400,000 applications.  If all else remained stable (which it will not), it would take 17 years to achieve this!  If the USPTO could increase the size of the examining corps overnight to 7,500, and halt the attrition (and then all else remained stable), the required period would drop to just a little over two years.  Viewed this way, a significant reduction in the backlog within the five-year time frame of the Strategic Plan seems quite achievable.

There is also huge scope to reduce the number RCEs filed.  If more applications can be finalised without the need for an RCE, it may not be unreasonable to achieve an additional 50,000 disposals per year.  Initiatives such as the First Action Interview Pilot Program are directed at precisely this issue.


Finally, we will look at the question of how many examiners are required in the longer term to ensure that the backlog, once tamed, does not return to its present state.  This is where our queueing model can provide particular insight.  The graph (right) shows the long-term average queueing delay (corresponding with pendency to a first action in our model) as a function of the ratio between arrival rate and service rate for the very simplest queueing model (known as the "M/M/1" queue, in queueing theory).  Obviously the USPTO is not a simple system, however the very large numbers involved cause a lot of complexities to "average out".  Additionally, the general form of the M/M/1 delay characteristic is common to many more complex queueing systems.  For these reasons, we expect that this vastly simplified model may provide useful insight.

The point to note about the graph above is that the queueing delay remains low until the arrival rate reaches about 90% of the service rate.  That is to say, once the backlog has been tamed, it may be necessary to maintain capacity at little more than 10% above the application rate in order to keep the backlog under control.  On this basis, an increase in the size of the examining corps by only 20%  from its present level is more than sufficient (unless there is a sudden jump in application numbers).  The numbers proposed in the Strategic Plan are therefore not only adequate, they leave some room for growth in the number of applications filed. 


Having looked at the numbers in some detail, we think that there is good reason to suppose that the goals set out in the Draft USPTO Strategic Plan are quite achievable, and to be optimistic that the current problems with backlog and pendency can be overcome within the next three to five years.

Much will depend upon the ability of the USPTO to attract and retain quality examiners.  However, given adequate budget, and the influence of a positive and creative Director in David Kappos, we are hopeful that this will be possible.

Scrapping the unpopular Final Rules, which directly addressed the problem but harmed the system in the process, was a positive move.  Other initiatives, such as the various PPH programs, the Green Technologies initiative, Project Exchange, multi-track examination and the First Action Interview Pilot Program, are all steps in the right direction that clearly address underlying factors that contribute to the backlog.

We set out on this statistical exercise with cynically low expectations that the data would reveal unrealistic claims and expectations.  We are glad to admit that we may have been wrong!


Except where other sources are indicated, data in this article is drawn from:


Ken Bower said...


I examined at the PTO for nearly 8 years ending in 2005. 18 hours for the average examination is absurd. Don't forget that the number of documents to be searched increases by over a million each year. The PTO does not keep up with reclassifying subclasses that are too large, many now with over 2,000 entries (a manageable sub class should have under 500 entries), and the huge number of citations required for accelerated examinations (hundreds of citations in one patent application) is making foreword and backward citation searches awesome in length.

You can not make a legitimate study without getting your hands dirty. How many examiners did you interview? How many uncompensated hours are they working? What is the affect of all of the senior examiners working at home? What is the affect of so many examiners with English as their second language?

Also, How many trainees are there. A trainee has a rough time doing one application a week. Then after a few months of frustration many of them leave.

I would really like to see the picture you paint after you get the data from below the surface.

Ken Bower

Unknown said...

Hi Ken

Thanks for your comment. Your input is much appreciated.

You are right - I did not interview any examiners. I think that was fairly clear from the article, as were the sources of information on which I relied. The 18-hour figure may not be correct, and I have nothing but Burk and Lemley's assertion to support it. They claim to have relied upon various sources, including their own contacts within the USPTO.

However, the number of examiners, the number of disposals, and the number of RCEs for FY 2009, are a matter of public record. The only conclusion that can be drawn from the published figures is that each examiner "serviced" (according to my definition), on average, about two applications per week. So if they were spending significantly more that 18 hours in total on each of those, then you would be right that they must be working overtime.

You seem a little offended by the article. Please believe that this was not my intent. In fact, I thought that I was painting a comparatively positive picture, although I have not doubt this is a matter of perspective.

I would welcome any further sources of reliable data.

Thanks again for your interest.


Post a Comment

Copyright © 2014
Creative Commons License
The Patentology Blog by Dr Mark A Summerfield is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Australia License.