EPO Performance 1 - application pendency times

The EPO has been telling all who will listen in recent years that its processing of applications is getting ever more efficient while quality is also improving. Given that quality is somewhat intangible and that the perception of different users on what constitutes quality can differ so much let’s leave that aside. Surely the efficiency claims can be verified beyond doubt?
Alas, the hard data and analysis has been somewhat sparse, leaving Merpel and the world at large in the position of having to rely on the EPO’s figures and claims with little opportunity to put these to the test.
Until now…
A valued reader who goes by the nom de plume of Archibald Calculus has been doing some number crunching. He has taken raw data of various types from the EPO, tabulated it, graphed it, and tortured it mercilessly until it gave up its essential truths.
Over the next few posts, Merpel will be passing on to her readers the results of Archibald’s invaluable analysis. Are pendency times improving measurably? Does an independent analysis back up the claims of efficiency made to the AC?
Read on and find out. 

Archibald takes up the story:

In decision T 0823/11 handed down 21 December 2015, EPO Board of Appeal 3.5.07 [1, 2] regarded examination proceedings lasting 12 years to be excessively long, and held that such a delay constitutes a substantial procedural violation. The grounds for the decision cited the Tyvik AS v Norway case adjudicated in 2013 by the European Court of Human Rights concerning excessively long national patent proceedings, which the judges in Strasbourg considered to have rendered meaningless the plaintiff's right of access to a court.
Proceedings before the EPO are notoriously lengthy. But just how long do they actually last? And what proportion of cases could potentially meet the criteria for a procedural violation?

The ministers of EPC member states defined objectives for the prosecution delay of European patent applications. According to the epi Information 4/2006, p. 126 these are:

14. The Paris Criteria (the time between filing and grant of European Patent Applications) […] arise from an intergovernmental conference in Paris on 24th and 25th June 1999 and include an invitation to the European Patent Organisation to undertake every possible effort to shorten procedures, so as to bring the average time it takes to grant a European patent down to three years, whilst maintaining the level of quality.

The preceding passage stems from a secondary source, and in view of the interpretation by the EPO, one can't say for sure whether the all-important “average” (emphasis added) qualifier had actually been specified by the ministers.

For instance, the 2003 activity report presented by President Ingo Kober (OJ EPO 2/2004, p. 79 ff) stated as follows:

The number of examination files not fulfilling the Paris criteria had slightly increased by 2.5% since the end of 2002, due to priority being allocated to search procedures. It had declined in six of the 14 industrial sectors. In 2003, the Office would meet the Paris criteria for one in five granted patents, up from one in seven for 2001.

This passage implies that the three-year objective is understood to be an absolute limit applying to individual files.

President Alain Pompidou appeared to adhere to the same interpretation as his predecessor in his keynote speech pronounced 5 November 2004 at the CIPA congress in London:

the number of files in the examination phase that do not meet the Paris criteria is decreasing.

Six years later, newly elected President Benoît Battistelli had an optimistic tone in his lecture held 8 November 2010 at University College London:

The average time taken to grant a patent in 2009 was 43 months, as in 2008. Thus, we are within striking reach of the Paris criteria of 36 months.

Document CA/88/14, “President's activities report”, provided on page 5/51 a few data points on the time required to grant a European patent for full years 2012 to 2014, and mentions that values presented are medians.

Six months later, the following edition of that report published as CA/44/15 stated that “[t]imeliness up to grant continues to improve”, and that “patents were published on average 42.3 months after the start of the European procedure”.

Information from the EPO on the duration of the procedure is rather piecemeal and seems confused as to what is being measured: mean, median, or proportion of files respecting some absolute criteria. But since CA/44/14 at last provides a definition of the manner in which the indicators are calculated, it becomes possible to verify them and understand what is being measured.



The figure above shows the result of this calculation, with the various data points gleaned superposed as green markers. In addition to the sources above, some data points were included from a presentation titled “Global Patent Warming” [sic] by EPO Controller Ciarán McGinley in 2008 at the IP Business Congress in Amsterdam. Slide 3/34 shows a bar chart labelled “Towards the Paris Criteria”, and provides a few interspersed numbers for the “Time to publication of patent”, without indicating whether these are means or medians, or details as to how they were obtained.
Applications can enter the European granting process either through direct filing under Art. 75 EPC (“EPdirect”), as a divisional application under Art. 76 EPC, or as a PCT application entering the regional phase Art. 153 EPC (“EuroPCT”). The dates used for the calculation were respectively the filing date accorded under Art. 90 EPC, the date of receipt of the divisional application documents, and the date of payment of the examination fee under Rule 159(1)(f) EPC. For the sake of completeness, applications can also be filed under Art. 61 EPC, but such applications are extremely rare except as answers to EQE problems, and thus can be safely neglected.

The lower curve, presenting the median delay of the events defined in CA/88/14, matches very well the data points gathered, under the proviso of defining a month to be 31 days, instead of the more logical average month 365.25/12=30.4375 days. At this point, one should realise that the EPO's habit of presenting occasional median value points tells very little about the shape of the  distribution of prosecution delays. The progressively worsening average value curve just above, and its widening gap from the median value suggest an increasing skewness of that age distribution.

Furthermore, the events chosen for starting the stopwatch are not equivalent for the three main types of applications, as each has a somewhat different time line, which also depends on factors including whether priority is claimed, whether the international application was filed in an EPO working language, and whether the EPO performed the international search. The resulting metric is therefore sensitive to applicant procedural choices, and it is difficult to discern their effect from variations in examination speed. To help understand this, the respective numerical proportions expressed in parts per thousand of the five main categories of the granted patents are shown in the lower part of the figure.

A different presentation is therefore necessary which provides more information on the shape of the prosecution delay distribution. The triggering event for the time measurement is the A publication in the European Patent Bulletin, which is more closely associated with the beginning of examination. This information is also more readily available than the events defined in CA/88/15. A “B” publication cannot occur before the “A” publication, but both events can be brought together under the provisions of Art. 93(2) EPC. The “A” publication is therefore a potential absolute zero for the measure of examination delay. The “A” publication is also the point where third parties are informed of the birth of provisional rights under Art. 67(1) EPC, and where the public's uncertainty as to the eventual ultimate scope of the rights begins. This starting point is also somewhat less sensitive to applicant choices and strategies, as the only major difference between the three major application types at the “A” publication date is whether an EPO-issued search report and opinion are available at that point.

What the next figure says isn't good news.


Despite the reassuring statements put out by almost all EPO Presidents in the last fifteen years^footnote including the present one, the delay to grant for European patents has worsened almost continually ever since the Office was founded.

The Office's practice of occasionally presenting median values for prosecution times, whether  deliberate or not, hides this situation.

One could take solace in that the above mean and average values presented in the above two figures remain roughly constant over time, but this is also a sign that emphasis is put on newer applications at the expense of older ones. And these values are not quite the full story, as will be seen in a further instalment.

Already about five percent of granted applications had been pending for more than 10 years, which is the kind of delay which the Boards of Appeal find objectionable. The uncertainty resulting from these delays also affects the right of third parties.

By comparison, twenty years ago, in 1996, the equivalent statistic was that five percent of granted cases had been pending for just six years; in 2006 the relevant value was 7.5 years. It’s now over 10 years. Looking at the 99th percentile line you can see that in both 1996 and 2006 there were effectively no cases that took over ten years to get to grant.

Similarly looking at the 75th percentile, one could say in 1996 that three-quarters of all granted cases were processed to grant in under 4.5 years. By 2006 it was taking 5 years for three-quarters of cases to be granted. Today it takes almost 6 years.

The above figures only show one type of procedural outcome, grants. In addition to this outcome, applications can exit the examination procedure in three different manners. It can be deemed withdrawn as a result by failing to fulfil some procedural obligation, most commonly to pay a renewal fee or answer a communication. This type of event is known by its mnemonic “ADWI”. It can also be explicitly withdrawn (“WDRA”), or refused (“REFU”), most commonly by the examining division under Art. 97(2).


The figure above shows the pendency distribution for all four outcomes, but also calculated from the legal filing date instead of from the “A” publication date. One should consider that the patent term is 20 years from filing, and that for the right to access to a court (or in the EPO context, a Board of Appeal) to be meaningful in the sense of the ECHR decision, there must be sufficient potential time left in the patent’s life. And this isn't even taking into account potential opposition and/or national revocation proceedings.

The population in the top percentiles is dominated by divisional applications, often undergoing appeals, which are typically filed several years into the life of the parent application. But if one removes these, the levels only improve by about a year or so, without affecting the general character of the graph.

The X-axis ends about six months ago, as it takes about that time for the legal effect date of most (>95%) negative outcomes to be recorded. Once again one can see that pendency times for the worst 5% have lengthened from almost 7 years (late 1995) to 9 years (late 2005) to nearly 12 years (late 2015). The 75th percentile similarly shows a steadily worsening trend: whereas three-quarters of cases were disposed of in 5 years in 1995, it took 6 years to dispose of 75% of cases in 2005 and by last year it was taking 7.5 years to do so.

---

Footnote ...with the exception of Alison Brimelow, who stated to IAM's Joff Wild back in 2008 that “the backlog won't be mastered”.^?