Analysis of Divisional Applications Trends in Light of Publicly Available Data
1 Introduction
The European Patent Office (EPO) introduced a multi-generation fee regime in 2014, following the abolition of time limits for filing divisional applications. The time limits were removed because they did not achieve their intended purpose and added unnecessary complexity to the European patent system.
After ten years, a renewed debate has emerged regarding potential misuse of divisional applications. Allegations of misuse are primarily linked to high-generation divisional applications and late filings, which extend the uncertainty for the third parties beyond the maximal duration of patent protection. Additionally, changes to legal regulations are reportedly being considered.
The EPO recently published a report on the processing of divisional applications (CA/PL 4/25Administrative Council documents are available at https://www.epo.org/en/about-us/transparency-portal#search). This report, prompted by discussions in the Committee on Patent Law and the Administrative Council, provides statistical data on divisional applications. Special focus was placed on two IPC classes – A61K (medicines) and C07D (organic compounds, including steroids).
The EPO has also implemented new measures to ensure the swift processing of divisional applications and increase legal certainty. These include issuing summons to oral proceedings as a first action in certain cases and publishing divisional applications as soon as possible after formalities checks.
The aim of this paper is to analyze overall trends in divisional applications without restricting the scope to specific IPC classes.
2 Data Source
This study relies on European Patent Bibliographic Data (EBD), which is publicly available via the EPO’s bulk data distribution service EBD Service. A detailed description of the data format is available at EBD – EPO Bibliographic Data Specifications. The dataset covers the years 2016-2024, allowing for a nine-year analysis.
EBD is driven by publications, meaning its data may not be directly comparable to EPO internal statistics. The analysis presented here predominantly consists in rather simple counting of publications A1/A2 (first publications) and B1 (granted patents), with additional analysis of corresponding dates and information regarding the generation of divisional applications.
3 Key Findings
3.1 Share of divisional applications
Numbers relating to the divisional patent applications should not be entirely put out of context of global numbers relating to patent applications and granted patents. Therefore, in figures 1, 2 and 7 regular (non-divisional) applications are given for reference.
3.1.1 Number of first publications of patent applications and fraction of divisional applications
The annual number of first publications (A1 or A2) of European patent applications, including the share of divisional applications, is shown in Fig. 1. While the data aligns rather closely with records from the European Publication ServerEuropean publication server available at https://data.epo.org/publication-server/?lg=en does not provide information about divisional applications but allows to get total numbers of annual publications., a discrepancy of approximately 2% is observed for the year 2020. Over the years, the fraction of divisional applications has varied, peaking at 8.6% in 2020 and then oscillating around 7-8% in more recent years. The increase, though noticeable, is not entirely consistent, making it difficult to assert a clear trend of rising divisional filings.
The most recent peak percentage of divisional patent applications among first publications occurred in 2023, likely driven by strategic responses to the Unified Patent Court (UPC) and unitary patent considerations.
3.1.2 Number of granted patents and fraction of patents granted on divisional patent applications
Fig. 2 presents the number of granted patents per year, distinguishing between patents granted on divisional applications and those granted on standard applications. While the data aligns rather closely with records from the European Publication ServerEuropean publication server available at https://data.epo.org/publication-server/?lg=en does not provide information about divisional applications but allows to get total numbers of annual publications., again discrepancy of approximately 2% is observed for the year 2020. The proportion of divisional applications among granted patents has increased, rising from 6.3% in 2016 to 8.2% in 2024. While this suggests a moderate increase, it remains relatively limited, and likely correlated or at least partly caused by broader changes in the legal framework, specifically the introduction of Unified Patent Court.
Neither Fig. 1 nor Fig. 2 indicate a rapid increase in the use of divisional patent applications.
3.2 Changes in the generation of division
Divisional patent applications can be parents to further divisional patent applications, forming longer or shorter chains of divisional applications. The order of generation of a division is defined as the number of parents in the chain. It is suggested that certain “misuses” of the divisional patent applications are correlated to the high generation of division. Fig. 3 provides histograms of first publications of divisional applications by generation of division.
An analysis of the generation of division is shown in Fig. 3, which displays histograms of first publications of divisional applications categorized by generation of division. Over time, a shift toward higher-generation divisional applications is noticeable, but the absolute numbers remain low. Instances of very high-generation divisional applications are exceptionally rare, typically amounting to only a few cases per year. The highest-generation divisional application observed, a 20th-generation divisional, was first published in 2019, and no subsequent divisional has approached this level.
A peak in first publications of divisional applications occurred in 2023, which can likely be attributed to the increased strategic use of divisional applications in response to the introduction of the Unified Patent Court (UPC). Applicants may have opted to file divisional applications to maintain flexibility between unitary patent protection and conventional validations. In fact, such strategies may still be in place as they were quite universally recommended prior to launch of the Unified Patent Court.
A similar analysis of granted divisional patents is presented in Fig. 4. While high-generation divisional patents have become somewhat more frequent in recent years, their numbers remain low. Between 2016 and 2020, no patents were granted on divisional applications of a generation higher than six. Between 2021 and 2024, patents on higher-generation divisional applications were observed, but the total count over four years amounted to just 11 patents. This small number does not provide strong justification for significant regulatory changes.
It is possible to observe that divisional applications of generation 7th and higher were not represented among granted patents in years 2016 – 2020. Higher-generation divisional patents became more frequent after 2021, but the total number remains small (11 patents granted on divisional patent applications of generation 7 or higher over four years). The numbers of patents granted on lower generation divisional show linear decreasing tendency in logarithmic scale, which translates to exponential decreasing. Generations higher than four are so infrequent that they could be dealt with individually.
It should be noted that chain of divisional applications need not be linear – divisional applications of all generations may have siblings. Accordingly, families of parent and divisional applications may be considerably larger than the mere generation of the youngest divisional in the chain of divisionals. Tracking of the families proved to be somewhat challenging and difficult to present in graphical forms but in years 2016-2025 at least some members of graphs of parent and divisional applications having 20+ elements have been found. For example EP08014611.1EP08014611.1, published as EP1985680, “Composition comprising a HFC-1234ze isomer”. is a first generation divisional and is a parent to 26 further divisional applications and member of family having 43 members. This example, however spectacular, is also very unique. Another one in a very different field of technology is EP16833045[EP16833045.4, published as EP3332846, “FRAUD DETECTION SYSTEM AT GAME PARLOR”.] being a parent to 24 divisional applications of different generations. As the members of those families were filed among more than 9 years ago it is difficult to take them into account in analyzing trends.
3.3 Timing of first publications after filing date
The time between the filing date and the first publication of divisional applications is analyzed in Fig. 5. The charts in Fig. 5 are based on annual data. For every first publication (A1/A2), publication of divisional application in a given year time between filing date and first publication date was computed. Histograms were prepared with bins having 1-year width. Every histogram comprises data of publications from one year. X-axis represent time bins, Y-axis represent number of publications within a bin. Histograms are represented in semi logarithmic scale (Y) to reveal trends in less frequent circumstances.
The most common timing for first publication is in the fifth year after filing, with the sixth year following closely behind. There is no evidence suggesting an increase in late-filed divisional applications. Although an extreme case was observed in 2024, where a divisional application was first published 25 years after filing, this was an isolated occurrenceIt was EP24186376.0 published as EP4417786A2, “A GEAR AND A FLUID MACHINE WITH A PAIR OF GEARS”..
To catch possible trends that might not have been visible on the charts, median and mean times between filing date and the publication date were computed (see Tab. 1). For reference data of non-divisional applications were added. Data for non-divisional applications are very stable, as there is little room for applicant decisions with respect to timing (perhaps early PCT entries). As far as divisional applications are concerned, it seems that both mean and average are decreasing so if any trend is to be observed, it would be to file divisional patent applications earlier.
It should be stressed that the data shows no indication that in general the applicants tend to file the divisional patents later in proceedings.
3.4 Timing of patent grants after filing date
Fig. 6 examines the distribution of granted patents based on the number of years elapsed since the filing date. Histograms presented in Fig. 6 represent an analysis of patents granted in years 2016-2024. Fig. 6Similar study was presented in CA/PL 11/22, however in CA/PL 11/22 all applications finalized were concerned (including applications refused, deemed withdrawn and withdrawn), therefore results are slightly different. is construed similarly to Fig. 5. Every histogram presents data of patents granted in one year. X-axis represent time between B1 publication and filing date with one year bins. Y-axis represent number of patents that were observed in certain bin. Histograms are represented in semi logarithmic scale (Y) to reveal trends in less frequent circumstances.
The data do not indicate any pattern of divisional patents being granted at increasingly older ages. Most divisional applications are granted between six and nine years after filing, which aligns with typical prosecution timelines. The occurrence of patents granted more than 20 years after filing (a scenario that may raise concerns) remains rare, with no observable increase.
Again, to detect possible subtle trends, average and median times in subsequent years were computed with non-divisional applications as reference. As shown in Tab. 2, neither median nor mean of the time from filing date to B1 publication date show tendency to increase in recent years.
Patents granted on divisional applications are older than the patents granted on non-divisional patents at the time of B1 by aproximately 4 years.
3.5 Processing time of divisional vs. non-divisional applications
The EPO report CA/PL 4/25 states that divisional applications are processed significantly faster than regular applications. As an example, in 2024, the average processing time for divisional applications was reported as 3.6 years, compared to 4.9 years for regular applications. The processing time cited by the EPO is measured from the date of receipt of an application to its finalization (whether by grant, refusal, or withdrawal).
This study, however, adopts a different approach to measuring processing time. Since EBD only provides information on published applications, it allows for tracking of applications with negative outcomes (refused, withdrawn, or deemed to be withdrawn) as well as those with positive outcomes (intention to grant). Alternatively, B publications provide information on B1 dates. Because it is not possible to recreate the data used by the EPO (which includes applications withdrawn before publication), for the sake of ease of comparison, this analysis of processing time has been limited to granted patents and is represented as the difference between the first publication date and the B1 date. The processing time reported in Fig. 7 as years between A1/A2 publication and B1 publication is expressed in years.
Fig. 7 presents histograms of processing times for granted patents in the years 2016-2024. The histograms are divided into divisional and non-divisional applications, with the number of patents in each histogram bin normalized by the total number of patents granted that year. Additionally, mean and median values are presented separately for divisional and non-divisional applications.
As shown in Fig. 7, in 2016 and 2017, the average processing times for divisional and non-divisional applications were quite similar, with non-divisional applications being processed in average slightly faster but having higher median time. The median processing time for divisional applications was higher and closer to the average, suggesting that extremely long prosecution times were less common for divisional applications. From 2018 onward, processing of divisional applications accelerated, with histograms shifting leftward, indicating shorter overall prosecution times. Especially in 2018, 2019, 2020, and 2021, it is evident that the bar corresponding to patents granted in the second year after the first publication date is the highest for divisional applications. In 2022-2024 there are two dominating bars corresponding to the 2nd and 3rd year after first publication, but the histograms remain more concentrated on the left the ones of 2016 and 2017.
By 2024, divisional applications were processed significantly faster than non-divisional applications, even if the difference was not as pronounced as suggested in the EPO report. This confirms that divisional applications are no longer inherently associated with prolonged examination times. The observed trend aligns with procedural adjustments at the EPO aimed at expediting the examination of divisional applications, particularly in cases where they closely resemble their parent applications.
4 Final observations
The share of divisional patent applications among the applications first published and among granted patents is quite stable in years 2016-2024, with observable increasing tendency.
The analysis reveals that while high-order divisional applications have become more common, they remain rare in absolute terms. Given the steady increase in overall application numbers and granted patents, it is natural that extreme cases will occasionally appear, but these do not constitute a widespread trend. Cases where the time of uncertainty is extended beyond the maximal duration of patent protection should be easily visible in charts Fig. 5 showing the time between the filing date (which is the same as filing date of the parent) and the first publications. Cases where divisional patent applications are first published close to 20 years from the filing date happen occasionally (e.g. EP24186376 first published in 25th year after the filing date) but are so infrequent that can easily be dealt with individually without any changes to rules applying to thousands of other divisional applications.
There is no indication that processing times of divisional patent applications are longer than regular or that divisional applications are becoming “older” at time of their first publication or at time of grant (B1 publication). The processing times (first publication to grant) for divisional applications have actually decreased in recent years. While the CA/PL 4/25 report suggests a significantly shorter time of processing of the divisional applications than non-divisional it seems that when counting from the first publication to grant the difference is lower, yet still divisional applications are processed faster.
Some effort was made to verify the presented data and statics across different EPO sources. However, it should be noted that EBD data do not correspond exactly to the data available in the European Publication Server, as highlighted in subsections 3.1.1 and 3.1.2 above. While care was taken to ensure accuracy, there remains possibility of mistakes or miscalculations. This article was prepared with the intention of making it as easy as possible for readers to verify the data using the same sources or other publicly available information.
5 Documents cited
CA/PL 4/25, CA/PL 11/22Administrative Council documents are available at https://www.epo.org/en/about-us/transparency-portal#search
