Content uploaded by Milda Aleknonytė-Resch
Author content
All content in this area was uploaded by Milda Aleknonytė-Resch on Apr 01, 2025
Content may be subject to copyright.
Making the Case for Process Analytics: A Use
Case in Court Proceedings
Milda Aleknonyt˙e-Resch1[0000−0003−0472−1262], Anna-Katharina
Dhungel2[0000−0003−0473−7790] Fabian Elsaeßer3, and Arvid
Lepsien1[0000−0002−8105−382X]
1Department of Computer Science, Kiel University, Kiel, Germany
{mar,ale}@cs.uni-kiel.de
2University of Lübeck, Lübeck, Germany
3Sozialgericht Kiel, Kiel, Germany
Abstract. Process mining and other forms of event data analytics have
shown to be valuable tools for supporting the management and execu-
tion of business processes. For this, process recordings in the form of
event logs are required. Digitalization efforts have led to an increased
availability of event data for many previously paper-based processes.
This has also inspired the extension of process analytics beyond classical
business processes. One example of this are judicial processes, which are
not necessarily bound by typical business constraints, but nonetheless
face issues related to, e.g., increasing efficiency or improving resource
allocation. Thus, the aim of this paper is to explore the usefulness of
process analytics to judicial processes using an exemplary use case from
a German social court. We show how event logs can be extracted from
digitalized court files and present an approach to identify bottlenecks
using these logs. The approach combines expert knowledge with data-
driven analysis and process mining. Using this approach, we are able to
both identify process inefficiencies and derive actionable insights for re-
ducing case durations, showing that process analytics is a promising tool
to facilitate the digitalization and optimization of judicial processes.
Keywords: Process mining ·Workflow inefficiencies ·Judicial processes
·Use case.
1 Introduction
With their process-first perspective, process mining and other forms of process
analytics have shown to be valuable tools in various applications involving busi-
ness processes, e.g., to identify bottlenecks in processes, detect deviations from
expected workflows, and support data-driven process optimization [18]. Typ-
ically, process analytics relies on event logs, which are recordings of business
processes extracted from business information systems. Due to the advancing
digitalization, an increasing number of processes are supported by information
systems. With this, more process-related event data is becoming available for
2 M. Aleknonyt˙e-Resch et al.
analysis [19]. While on the one hand, this enables gaining a deeper perspective
on business processes, on the other hand it has also inspired extending the scope
of process analytics beyond the typical business setting [2].
Judicial processes present a promising application, even though they differ
fundamentally from business processes. While the latter aim to generate profit
and offer design flexibility, judicial processes are bound by procedural law and
lack a profit-driven objective. In other aspects however, business and judicial
processes appear very similar. Both generally aim to add value to their organi-
zation [9], both are concerned with the work of people inside these organizations,
and most importantly, both face similar questions: What are the possible causes
of delays in the process? Are there bottlenecks? How can we improve the duration
of cases? How should resources (machines, workers, judges, ...) be allocated?
In recent years, courts started transitioning to electronic case management,
which enables the extraction of structured event logs [5]. With digital case files
becoming mandatory in many countries in the near future (e.g., in January 2026
in all of Germany), an even larger availability of structured event data can be
expected. Given that process analytics has successfully been used to address
the above-mentioned questions in business contexts, exploring its application to
judicial processes promises to provide a valuable set of tools for this domain
while simultaneously extending the scope of process analytics.
Consequently, the goal of this paper is to investigate the applicability of
process analytics to judicial processes. For this, we present a case study on the
handling of lawsuits in a German social court, where we extracted an event
log, analyzed it with an approach combining data-driven findings and expert
knowledge, and generated actionable insights for process improvement. Besides
demonstrating the applicability and usefulness of process analytics and process
mining for judicial processes, this paper also gives insights on how the analysis
of judicial processes can be approached successfully. Additionally, the real-world
event log extracted from the court data is provided for future research.
The remainder of the paper is structured as follows. In Section 2, the method-
ology and context of the case study are summarized. Section 3 describes the
dataset used in our case study and outlines our exploratory analysis approach.
Section 4 presents the application of our approach to the case study data. This
paper ends with Section 5 where we summarize our findings, discuss limitations
and lessons learned from the case study, and present directions for future re-
search.
2 Context and Planning
2.1 Case study methodology
The case study was conducted based on the widely-used PM2methodology [10].
PM2divides process mining projects into six subsequent phases that may also
be repeated in an iterative process if deemed necessary. Briefly summarized in
terms of the PM2phases, the case study was conducted as follows: After the
Making the Case for Process Analytics: A Use Case in Court Proceedings 3
Hearing
required
Inform
Passive Party
(Defendant)
Examine
Case File
Direct
Ruling
Invite Parties
to Court
Hearing
Call Parties
to Court
Court
Date
Hearings
finished
Settlement
Write
Verdict
Re-schedule
Lawsuit filed
No hearing
required
Send verdict
Request
Institution
Request
Other
Request
Expert
Witness
Request
Passive
Party
Request Active
Party
Investigation
Further investigation needed
Continue
case
Adjourn-
ment
Stage 3: Trial
Stage 2: Review and
Case Evaluation
Stage 1: Case
Initiation
Fig. 1. BPMN model giving a high-level overview of the analyzed court process.
planning phase (Phase 1), which involved initial discussions to establish a basic
understanding of the process and set the analysis goals, as well as a review of
related literature, we extracted (Phase 2) and processed (Phase 3) an event log of
the lawsuit handling process. Then, an exploratory analysis approach combining
expert opinion with data-driven insights was designed and applied to the event
log (Phase 4). The results of the analysis were evaluated and discussed with do-
main experts (Phase 5), resulting in actionable insights for process improvement
(Phase 6). At the time of this paper’s submission, Phase 6 had just begun, with
domain experts leading its implementation to integrate the identified process
improvements into judicial practice. In the following sections, the actions taken
in each phase of the case study and their results are presented.
2.2 Anatomy of the lawsuit handling process
The planning of the project (Phase 1) was initiated via a series of interviews
and discussions with domain experts, with the aim of (1) selecting the specific
process to be analyzed, (2) achieving a general understanding of the process, and
(3) setting the goals of the analysis. From these discussions, it was decided that
the project should focus on the process by which lawsuits are handled in the
court, with the goal of identifying activities and other factors associated with
process inefficiencies and increased case durations.
Figure 1 presents an abstract overview of the analyzed process. Generally,
the process unfolds in three stages. The first stage, case initiation, encompasses
all actions taken before an official lawsuit is filed, e.g., the plaintiff preparing a
lawsuit individually or with assistance from a lawyer or court clerk. As this stage
is executed by the plaintiff according to their personal decisions and independent
of how the court handles its lawsuits, it is considered out of scope for our case
study. Instead, the paper will focus on the subsequent stages, which are initiated
when an official lawsuit is submitted and then executed by the court and its
judges. In stage 2, review and case evaluation, the judge informs the defendant
of the lawsuit, reviews the case and performs an investigation to collect the
4 M. Aleknonyt˙e-Resch et al.
information necessary for a ruling. While in other fields of law, e.g., civil law,
evidence is typically provided by the opposing parties, in social law the judges
themselves are responsible for gathering all required evidence. This may involve
requesting additional information from institutions, the active (plaintiff) and
passive (defendant) parties, expert witnesses, or other sources. Based on the
investigation, the judge determines whether a hearing is necessary. If no hearing
is required, a direct ruling is issued, and the verdict is sent to the parties. If a
hearing is required, the trial (stage 3) is initiated by inviting all relevant parties
and scheduling a court date. At the court hearing, the case can be resolved in
multiple ways. A verdict may be issued, leading to either acceptance or appeal
by the parties. Alternatively, the parties may reach a settlement, or the court
may decide to adjourn the proceedings, potentially returning to the investigation
phase for further evidence gathering. Furthermore, the plaintiff can withdraw
their lawsuit at any point in the process.
While the trial stage is fairly structured and based on strict procedural rules,
the investigation stage is highly complex and heterogeneous, mainly due to two
reasons. Firstly, the constitutionally guaranteed independence of the judiciary
gives judges a great deal of freedom in the way they conduct investigations. Sec-
ondly, each investigation requires case-specific steps, which might, for instance,
involve concurrently and repeatedly identifying and calling witnesses, request-
ing expert opinions and coordinating with institutions. This poses a challenge
for process mining techniques, which work best with structured processes [8].
However, process mining might still produce objective insights about the struc-
ture of the process, and together with other process analytics techniques identify
patterns in procedural delays and bottlenecks that may otherwise go unnoticed.
2.3 Related analyses and approaches
To finalize the planning phase, related literature was screened for analyses con-
ducted on similar data, and for related analysis approaches. The literature
screening showed that process analytics techniques have previously been applied
in several case studies of judicial processes. These case studies were concerned
with civil proceedings from Italy and Brazil, where process mining and other
data-driven techniques have been applied to assess the impact of digitalization
on court efficacy [5], to analyze the evolution of processes over time [4], to iden-
tify activities significantly impacting process duration [4,16,19], to detect sources
of process delays [15,3], to engineer features for machine learning [20], to predict
the remaining duration of ongoing cases, e.g., for operational support [15,3], and
to communicate analysis results to stakeholders [19]. To the best of our knowl-
edge, there are no existing case studies that apply process mining to German
court proceedings or specifically to social courts.
Identifying process inefficiencies or performance anomalies and their causes,
which are also commonly called bottlenecks [12], was a common goal in papers
analyzing judicial processes. Various techniques exist to detect bottlenecks, for
instance using statistics or machine learning [12]. With its process-first perspec-
tive, process mining has been shown to be a viable method for bottleneck analy-
Making the Case for Process Analytics: A Use Case in Court Proceedings 5
sis. To this end, Bemthuis et al. [1] review approaches to detecting and predicting
bottlenecks as well as recommending improvements using process mining. Based
on this, Piest et al. [17] present a method for handling bottlenecks using process
mining. In most existing approaches (see [1,17]), process mining is generally used
as a supportive tool, e.g., discovered process models annotated with throughput
times are inspected manually to identify bottlenecks [1,19]. Techniques using
process mining for the automated detection of bottlenecks primarily exist for
specific applications, for instance, in manufacturing [11].
3 Materials and Methods
3.1 Dataset
Following the PM2methodology [10], Phase 2 (extraction) involved retrieving
data from the court system. Data was extracted in PDF format with personal
information redacted to ensure privacy. The raw dataset consisted of 260 cases
from three chambers within a single German social law court. The data origi-
nates from a single judge, who typically oversees five to six chambers, meaning
that this dataset represents only a subset of the judge’s total caseload. Optical
Character Recognition (OCR) was used to extract the document text, which was
organized into an event log according to the tabular structure of the documents.
In the dataset, a single timestamp is recorded for each activity, commonly indi-
cating only the date of occurrence rather than a precise timestamp. This limits
the granularity of time-based analyses and the accuracy of calculated activity
durations. Depending on the analysis goals, the occurrence of two activities on
the same date might also imply uncertainty about their order caused by these
imprecisions [13], which might in turn require special handling (e.g., [14]). As
the analysis focuses on the overall durations of cases, which typically range from
multiple months to years, the impact of the timestamp imprecisions was negli-
gible in our use case.
After extraction, the event log was further processed in consultation with do-
main experts to ensure anonymity, remove noise, and raise it to an abstraction
level appropriate for analysis (Phase 3). All remaining personal identifiers, such
as expert witness names, were removed from the log to ensure anonymity. Addi-
tionally, timestamps were systematically perturbed to further enhance data pri-
vacy. Originally, the event log contained 22,664 recorded events and 290 unique
activities. Activities that were extremely rare (i.e., occurring fewer than 30 times)
were excluded to focus on frequently observed procedural steps. Furthermore,
the domain experts reviewed the list of unique activity labels, based on which
similar activities were merged, and terminology was standardized across cases.
The refinement of the activity labels reduced the number of unique activities to
59. Finally, duplicate events were removed. These steps collectively reduced the
dataset to 19,947 events. The anonymized and processed dataset with 260 cases,
19,947 events and 59 unique activities can be found in this repository.4
4Link will be added when manuscript is published.
6 M. Aleknonyt˙e-Resch et al.
DATA-DRIVEN
INITIATION
EXPERT OPINION
INITIATION
3. Data Processing 4. Mining & Analysis 5. Evaluation
Enhanced
models
Expert
Interviews
Data Trans-
formation
Initial
Examination
Regression
Split Log &
Discover
Models
Descriptive
Statistics
Process
Insights
Correlation
Interpretation
& Discussion
Shortlist possible
Bottlenecks
Bottleneck
candidates
indicates steps where
domain experts were involved
Fig. 2. Exploratory analysis approach combining expert opinions and statistical anal-
ysis, aligned to the phases of the PM2methodology.
3.2 Exploratory Analysis Approach
To address the question of why some cases take significantly longer than others,
we designed an exploratory approach for analyzing the event log that encom-
passes phases 3-5 of PM2. While bottlenecks are a subset of the factors contribut-
ing to case duration, some cases may inherently require more time due to their
complexity or legal requirements, which cannot be optimized. To systematically
investigate the factors influencing case duration, we employed a two-fold initi-
ation approach to further processing the log for analysis that integrates (A) a
data-driven statistical analysis with (B) domain expert knowledge (see Figure 2).
Data Processing The data-driven initiation (A) involved transforming the
event log into a structured dataset suitable for correlation and regression anal-
ysis. To achieve this, we calculated key features for each case, including case
duration, the number of unique events per case, the total number of events per
case, and a dummy variable indicating the presence of each event type. A cor-
relation analysis was then conducted across all variables to identify potential
relationships with case duration. From this analysis, we selected the variables
with the lowest p-values while ensuring that highly intercorrelated variables were
excluded to avoid multicollinearity in the regression model. Meanwhile, the ex-
pert opinion initiation (B) involved interviews with judges (domain experts)
from multiple different social courts. Through these interviews, we identified
variables that judges perceive as bottlenecks. These variables were pre-screened
with descriptive statistics to estimate their importance.
Mining & Analysis In the mining & analysis phase, we built a regression model
using the variables from the correlation analysis that had low p-values and did
not exhibit high intercorrelation. A backward elimination procedure was then
applied to identify the most statistically significant predictors of case duration.
This regression analysis provided a shortlist of possible bottlenecks. This list
was merged with the results of the expert opinion initiation into a combined
Making the Case for Process Analytics: A Use Case in Court Proceedings 7
set of bottleneck candidates. Data subsets were then created by splitting cases
into groups with and without these bottlenecks, allowing us to assess whether
specific procedural patterns or case attributes were strongly associated with
extended case durations. Additionally, the dataset was segmented based on case
duration, deriving process models for the 20% fastest cases, the 20% slowest
cases, and the 20% of cases closest to the mean duration. Process models were
built from these subsets using Disco (Fuzzy Miner). Only cases with clear start
and endpoints were included, as defined by the domain experts. Additionally, we
gathered descriptive statistics on the process models, e.g. mean and median case
durations, and the total number of unique activities. An initial interpretation of
the analysis results was conducted without the domain experts to prepare the
subsequent discussions.
Evaluation The evaluation phase involved comparing the descriptive statistics
and the derived process models to assess differences between cases affected by
bottlenecks and those that were not, aiming to identify any structural differences
that may explain prolonged case durations. Additionally, key metrics such as case
duration, number of events per case, number of unique activities, and number
of directives were analyzed to quantify the impact of identified bottlenecks. The
process model visualizations and statistical findings were then discussed with
domain experts to validate the results and ensure their practical relevance. This
expert feedback played a crucial role in interpreting the findings and refining the
analysis. Finally, the evaluation led to defining potential process improvements
and support measures to optimize case handling and reduce inefficiencies in court
proceedings.
4 Results
Data Processing Our initial interviews with five judges from various social
courts identified several perceived bottlenecks that could contribute to extended
case durations. These included the involvement of expert witnesses, the need to
send reminders, the request for medical findings, treatment reports, and whether
a court date was required.
Regarding the inclusion of expert witnesses, it was argued that it introduces
a structured delay. While expert witnesses are typically given a fixed response
time to submit their report, in some cases, they require additional documents,
which can further extend their response time. Similarly, when reminders need
to be sent to parties, judges noted that this often indicates a pattern of late
responses, meaning that the involved party is more likely to only respond at the
last possible moment throughout the entire process.
Additionally, the need to request medical findings and treatment reports
mean that additional documents must be collected, which involves multiple ac-
tors and can further delay proceedings. This is particularly tedious because the
judge must wait for the requested information and then revisit the case file once
the documents arrive. Finally, not all cases require a court date. Judges believed
8 M. Aleknonyt˙e-Resch et al.
Fig. 3. Correlation with the case duration and intercorrelation of selected activities.
Numerical values represent correlation coefficients.
that cases where a verdict could be written immediately should be resolved
faster, as Stage 3 (the trial stage, see Figure 2) would be significantly shorter
compared to cases where a hearing is necessary.
In the data-driven initiation branch, 22 out of 65 variables exhibited a Pear-
son correlation p-value of less than 8×10−4, indicating a possibly statistically
significant relationship with case duration. Since this analysis is of exploratory
nature, no statistical significance threshold was set. The correlation strengths
between case duration and these variables, along with their intercorrelations,
are visualized in Figure 3. A number of strong intercorrelations between certain
variables can be observed. This suggests procedural dependencies. For exam-
ple, a perfect correlation was found between "Request for Medical Findings and
Treatment Report" and the subsequent submission named report. This is an
expected outcome, as such a report is always submitted when the judge requests
it, making this dependency structural.
Making the Case for Process Analytics: A Use Case in Court Proceedings 9
The highest correlation coefficients with case duration were observed for the
total number of events (0.69), the number of unique events (0.48), and the pres-
ence of an expert witness report (0.44). The strong correlation between the
number of events and the number of unique events is expected, as cases with
a higher number of distinct procedural steps tend to be more complex, requir-
ing additional interactions and decisions. Similarly, the presence of an expert
witness report is indicative of greater case complexity. Since obtaining such a
report necessitates an external review, it inherently extends case duration due
to dependencies on third-party availability and evaluation time.
In contrast, the lowest correlation coefficients with case duration were found
for case withdrawal (-0.25), preliminary lawsuit submission (-0.26), and chamber
number (-0.30). The negative correlation of withdrawals with case duration is
intuitive, as cases that are withdrawn do not require a formal verdict, signifi-
cantly shortening their processing time. Interestingly, withdrawal correlates with
preliminary lawsuit submission, suggesting that cases involving a preliminary
lawsuit submission (often submitted by associations) frequently lacked sufficient
merit and were later withdrawn. The correlation with chamber number is likely
a spurious correlation, meaning it does not reflect a true causal relationship but
rather an incidental statistical association.
Interestingly, the activity send reminders shows no correlation with any other
activity but exhibit a correlation coefficient of 0.28 and p-value of 6×10−6with
case duration. This suggests that once a reminder has to be sent for a party
to respond, the entire process is slower. This could indicate that certain parties
systematically delay and/or are slow with their responses, prolonging the overall
case duration. Such findings highlight that bottlenecks may not always stem
from the procedural structure itself, but from the behavior of involved parties.
Mining & Analysis To identify key factors influencing case duration, 11
variables with low intercorrelation were selected from the correlation analysis
and included in a regression model. After applying a backward elimination pro-
cedure, 7 variables remained nominally significant and thus identified as possible
bottlenecks. The regression results, including estimates, standard errors, and p-
values, are presented in Table 1. The estimate values can be interpreted as the
number of days a given factor alters the case duration on average.
The results indicate that the inclusion of an assessment document in a case
file is associated with a 198-day increase in case duration. Since assessment doc-
uments can be submitted at the beginning of the process alongside the lawsuit,
this may indicate larger, more complex cases or cases that involve legal repre-
sentation. Similarly, if an expert witness report is required, the case duration
is extended by 121 days, reflecting the additional time needed for the external
assessment.
Other factors contributing to longer case durations include individual or-
ders (+78 days) and attachments from the plaintiff (+67 days), both of which
suggest more extensive case documentation and procedural steps. Conversely,
certain factors were associated with shorter case durations. The submission of a
preliminary lawsuit reduces case duration by an average of 96 days, likely due
10 M. Aleknonyt˙e-Resch et al.
Table 1. Regression after backward procedure. Adjusted R2: 0.399, p-value of F-
statistic: 2.2×10−16.
Estimate Std. Error Pr(>|t|)
(Intercept) 47.338 88.470 0.593 069
No. Unique Events 15.075 4.259 0.000 48
Assessment document 198.371 38.843 6 ×10−7
Individual Order 77.958 32.076 0.015 78
Expert Witness Report 120.935 48.583 0.013 45
Attachment from Plaintiff 68.671 32.891 0.037 82
Submission of Preliminary Lawsuit −96.383 29.386 0.001 19
Withdrawal −66.743 30.419 0.029 15
to the streamlined preparation and experience of social welfare organizations in
handling such cases. Finally, case withdrawals shorten the process by 67 days,
which is expected as withdrawn cases may bypass multiple procedural steps and
do not require a final judgment.
Process models were derived from subsets of cases, where the subsets were de-
fined based on the variables that remained nominally significant in the regression
analysis above. In order to examine the impact of these factors on procedural
variations, process models were visualized separately for cases with and with-
out each significant variable. To ensure meaningful process visualization, a filter
was applied in agreement with domain experts, trimming the cases to sequences
beginning with a (preliminary) lawsuit filing and ending in either a court rul-
ing, direct court ruling, or verdict, reducing the dataset to 17,819 events. Due
to space constraints, we present only the process model for cases without ex-
pert witnesses. As shown in Figure 4, it is clear and easy to interpret, unlike
the complex and entangled spaghetti model observed when expert witnesses are
involved.
Hereafter, the discussed results are presented in Table 2, which provides a
detailed comparison of case durations, event counts, and judicial involvement
across different case subgroups. Due to space constraints, we focus on subsets
based on the presence of expert witnesses, the presence of assessment documents,
and case duration categories (fastest, slowest, and around the mean). Across all
subgroups, the mean and median case durations are very similar, suggesting that
the data within each subgroup is approximately normally distributed, meaning
there is no strong indication of skewness or extreme outliers.
The presence of an expert witness has a substantial impact on case duration.
Cases without expert witnesses have a mean duration of 8.3 months, whereas
cases with expert witnesses take significantly longer, averaging 17.4 months.
This discrepancy is further reflected in process complexity: cases without expert
witnesses have fewer unique activities (41 vs. 56), fewer total events (39 vs. 80),
and fewer directives per case (8.46 vs. 16.48). This suggests two key conclusions:
first, cases requiring expert witnesses tend to be more complex, as they involve
Making the Case for Process Analytics: A Use Case in Court Proceedings 11
Fig. 4. Process model visualizing mean duration of the log subset containing only cases
not involving expert witnesses.
additional procedural steps and evidence review. However, the higher number of
directives per case in cases with expert witnesses indicates that judges may need
to conduct further investigations even after contacting an expert witness. This
suggests a potential inefficiency, confirmed by domain experts: if judges could
gather all necessary information for expert witnesses earlier in the process, it
might be possible to reduce delays and streamline case progression.
It can be seen that the presence of assessment documents correlates with
longer case durations. Cases with assessment documents have a mean duration
of 25.2 months, whereas those without take 17.2 months on average. Addition-
ally, cases with assessment documents involve more events per case (90 vs. 74)
and more directives per case (18.03 vs. 14.47) as seen in Table 2. Assessment
documents serve as an indirect indicator of case complexity, and unlike expert
witness involvement, this complexity is already apparent in Phase 1 (when the
lawsuit is filed). According to domain experts, the underlying reason for in-
creased complexity can stem either from the intrinsic nature of the case itself or
from the involvement of a lawyer.
Subsequently, a comparison of the fastest 20% of cases, cases around the
mean duration, and the slowest 20% of cases reveals significant differences in
process length and complexity. The mean case duration for the fastest cases is
6.2 months, whereas the slowest cases take an average of 23.6 months, nearly four
times longer. A key finding is that slowest cases have more than twice as many
events per case (102 vs. 41) compared to the fastest cases. This suggests that the
12 M. Aleknonyt˙e-Resch et al.
Table 2. Descriptive statistics for case subgroups.
20% Cases that are Expert
Witness
Assessment
Documents
Total Fastest Around
Mean Slowest Yes No Yes No
Number of Cases 254 51 53 50 197 57 38 222
Mean Events per Case 70 41 73 102 80 39 90 74
Number of Activities 59 50 53 51 56 41 51 59
Median Case
Duration (months) 14.8 7 16 25.7 16.6 7.6 25 16.6
Mean Case
Duration (months) 15.3 6.2 16.2 23.6 17.4 8.3 25.2 17.2
directives per Case 14.64 8.28 15.36 21.51 16.48 8.46 18.03 14.47
primary reason for extended case duration is not necessarily waiting times, but
rather additional procedural steps that must be taken within the legal process.
Further supporting this observation is the number of directives per case, which
reflects the level of judicial involvement. The fastest cases require significantly
fewer directives per case (8.28), roughly half of the average cases (15.36), whereas
slowest cases require 21.51 directives per case. This indicates that more complex
cases demand substantially more input from judges. Furthermore, it suggests
that identifying cases likely to require additional procedural steps early in the
process may help courts allocate resources more effectively, potentially mitigating
extensively long case durations.
Evaluation of Use Case In Phase 5, the results were discussed with do-
main experts to validate their practical relevance and interpretability. The judges
confirmed that all the findings aligned with their professional experience and ex-
pectations, reinforcing the credibility of the analysis. The most interesting results
for the domain experts were the strong effects of expert witnesses and assess-
ment documents. While they had already observed that cases involving expert
witnesses and assessment documents tended to take longer, the magnitude of
these effects (123 days longer for expert witnesses and 197 days longer for as-
sessment documents) was particularly impressive. Having a quantified estimate
of these effects provided them with a more concrete understanding of proce-
dural delays. Additionally, the discussion helped clarify causal relationships, as
some insights only became apparent after analyzing the data. For example, it
was only through the exploratory analysis that the presence of assessment docu-
ments was recognized as a strong early indicator of case complexity. This means
that courts could potentially use this information as a predictive signal in Phase
1 to anticipate longer case durations and allocate resources accordingly.
Together with domain experts, we could derive actionable insights for im-
proving court proceedings and reducing bottlenecks. These actionable insights
include early identification of complex cases, particularly those involving expert
Making the Case for Process Analytics: A Use Case in Court Proceedings 13
witness reports or assessment documents, which significantly prolong case du-
rations. By flagging such cases at the beginning of the process, courts could
allocate resources more effectively and plan for necessary procedural steps in
advance. Additionally, since cases requiring multiple directives, expert witnesses
and extensive documentation handling tend to take longer, process efficiency
could be improved by streamlining information requests and ensuring that all
necessary documents are gathered as early as possible.
Another insight is that cases where reminders are sent often experience sys-
tematic delays, indicating that some parties consistently respond at the last
possible moment. Courts could mitigate this by implementing stricter follow-up
mechanisms or procedural adjustments to reduce unnecessary waiting times.
5 Discussion and Conclusion
In this paper, we explored the applicability of process mining and process ana-
lytics in judicial processes by conducting a case study in a German social court.
Our approach combined data-driven statistical analysis with expert knowledge
to identify bottlenecks and variations in case duration. By extracting and an-
alyzing event logs from digitalized court files, we demonstrated that process
analytics can provide valuable insights into court proceedings. Our use case spe-
cific contributions include identifying key factors that extend case duration, such
as expert witness involvement and assessment documents, and providing action-
able insights to optimize judicial workflows. Through discussions with domain
experts, we validated our findings and highlighted the potential of process mining
to enhance transparency and efficiency in legal decision-making.
In our use case, we demonstrated that process analytics is both feasible and
valuable for analyzing court proceedings data. By applying BPM techniques us-
ing our exploratory data analysis approach, we were able to gain first insights
into the judicial process. Our findings were not only data-driven but also val-
idated through discussions with judges, reinforcing the practical relevance of
process analytics in the judicial domain. This collaborative approach helped to
convince legal professionals of the potential benefits of process analytics, show-
ing that process mining can enhance transparency, explain variability in case
durations and identify bottlenecks in court proceedings.
Our analysis compared perceived bottlenecks (Approach B) with actual bot-
tlenecks (Approach A), revealing that while the identified delays were expected,
their magnitude was surprising. Cases involving expert witnesses and assessment
documents extended case durations by 121 days and 198 days, respectively. Al-
though assigning an expert witness is not a direct bottleneck, it alters the process
flow, often prolonging cases unnecessarily. Additionally, individual orders and at-
tachments from plaintiffs extended cases by over two months, an effect that was
previously underestimated by domain experts.
By comparing the slowest and fastest cases, we found that process complex-
ity, measured by the number of events, directives, and unique activities, could
be a stronger determinant of case duration than waiting times alone. These
14 M. Aleknonyt˙e-Resch et al.
findings, validated by domain experts, provided valuable insights into procedu-
ral inefficiencies and confirmed that certain case attributes (such as assessment
documents) can serve as early indicators of complexity.
In our use case, the actionable insights derived from the analysis have proven
highly valuable for domain experts. The results resonated well with judicial prac-
titioners, who appreciated the clarity the analysis provided, especially regarding
the versatility of Stage 2. While Stages 1 and 3 follow well-defined, structured
workflows, Stage 2 is notably dynamic and subject to individual judicial dis-
cretion (particularly when expert witnesses are involved) resulting in a lack of
a “happy path.” Another key insight was the identification of specific activi-
ties, such as the inclusion of additional and assessment documents at the end of
Stage 1 or the beginning of Stage 2, which are strong indicators of potential case
prolongation and complexity. This finding suggests that non-judicially-binding
guidelines could be developed to (1) streamline the process when expert witnesses
are involved and (2) help judges recognize early signals of increased complexity,
thereby enabling them to better plan and allocate their resources. Overall, these
insights not only affirm that BPM can be effectively applied to court proceed-
ings but also provide practical, data-driven strategies that can support judicial
decision-making and workflow optimization.
A key limitation of this study is that the use case stems from one social law
court in Germany, focusing on chambers handling specific types of cases from
the same judge. While the insights gained are valuable for this context, judi-
cial processes can vary significantly across different courts, regions, and legal
systems. However, our exploratory approach is not limited to this specific use
case. Given similar structured event log data from other courts or even different
countries, the methodology could be replicated and adapted to derive further
insights. Furthermore, the event log available from the court system contained
only start dates for events. While certain activities, such as sending invitations,
are instantaneous and the most important aspect of the event log in court pro-
ceedings is the chronological sequence of the process, the duration of events could
provide deeper insights into bottlenecks and inefficiencies.
Looking to the future, judges expect that new technologies will shorten court
proceedings [7]. The introduction of the electronic case file in Germany, in con-
junction with BPM, has the potential to make proceedings more efficient and
to identify process slowdowns at an early stage. Future work should focus on
analyzing directive types, as different directives may impact case duration dif-
ferently. LLMs could help extract and categorize directive content from PDF
files, enabling a more detailed process analysis. Extraction of further informa-
tion e.g. duration of events by means of meta data would increase the quality
of the event logs. Additionally, reducing complexity in process mining for court
proceedings is essential, as legal processes are highly heterogeneous. Simplifying
event logs, grouping similar process variants, and filtering non-essential steps
would make BPM more effective and accessible for judicial analysis. Addressing
these challenges will further optimize court processes, improve case management,
and enhance transparency and interpretability [6].
Making the Case for Process Analytics: A Use Case in Court Proceedings 15
References
1. Bemthuis, R., Van Slooten, N., Arachchige, J., Piest, J., Bukhsh, F.: A Classi-
fication of Process Mining Bottleneck Analysis Techniques for Operational Sup-
port:. In: Proceedings of the 18th International Conference on e-Business. pp.
127–135. SCITEPRESS, Online Streaming (2021). https://doi.org/10.5220/
0010578601270135
2. Brzychczy, E., Aleknonyt˙e-Resch, M., Janssen, D., Koschmider, A.: Process mining
on sensor data: a review of related works. Knowledge and Information Systems pp.
1–34 (2025)
3. Campi, A., Ceri, S., Dilettis, M., Pernici, B.: Variants Analysis in Judicial Tri-
als: Challenges and Initial Results. In: ECML PKDD 2023 Workshops. CCIS,
vol. 2133, pp. 425–438. Springer, Cham (2025). https://doi.org/10.1007/
978-3-031-74630- 7_30
4. Caponecchia, V., D’Agostino, B., Comandè, G., et al.: Towards visualizing and
analysing legal proceedings with process mining. In: 1st International Workshop on
Processes, Laws and Compliance. CEUR Workshop Proceedings, vol. 3850. CEUR-
WS.org, Lyngby, Denmark (2024), https://ceur-ws.org/Vol-3850/paper5.pdf
5. Castelliano, C., Grajzl, P., Watanabe, E.: Does electronic case-processing enhance
court efficacy? new quantitative evidence. Gov. Inf. Q. 40(4), 101861 (2023).
https://doi.org/10.1016/j.giq.2023.101861
6. Dhungel, A.K., Beute, E.: AI Systems in the Judiciary: Amicus Curiae? Interviews
with Judges on Acceptance and Potential Use of Intelligent Algorithms. In: ECIS
2024 Proceedings. AIS, Paphos, Cyprus (2024)
7. Dhungel, A.K., Heine, M.: Cui bono? Judicial decision-making in the era of AI: A
qualitative study on the expectations of judges in Germany. TATuP 33(1), 14–20
(2024). https://doi.org/10.14512/tatup.33.1.14
8. Diamantini, C., Genga, L., Potena, D.: Behavioral process mining for unstruc-
tured processes. J. Intell. Inf. Syst. 47(1), 5–32 (2016). https://doi.org/10.1007/
s10844-016-0394-7
9. Dumas, M., Rosa, M.L., Mendling, J., Reijers, H.A.: Fundamentals of Business
Process Management. Springer Berlin, Heidelberg (2018). https://doi.org/10.
1007/978-3-662-56509- 4
10. van Eck, M.L., Lu, X., Leemans, S.J.J., van der Aalst, W.M.P.: PM2: A pro-
cess mining project methodology. In: Advanced Information Systems Engineering
(CAiSE 2015). pp. 297–313. Springer, Cham (2015). https://doi.org/10.1007/
978-3-319-19069- 3_19
11. Fang, Z., Yu, C.: Bottleneck Mining: A data-driven bottleneck identification
method via process mining in manufacturing systems. In: 20th International Con-
ference on Automation Science and Engineering (CASE). pp. 1626–1631. IEEE,
Bari, Italy (2024). https://doi.org/10.1109/CASE59546.2024.10711833
12. Ibidunmoye, O., Hernández-Rodriguez, F., Elmroth, E.: Performance Anomaly De-
tection and Bottleneck Identification. ACM Comput. Surv. 48(1), 1–35 (Sep 2015).
https://doi.org/10.1145/2791120
13. Leemans, S.J.J., Van Zelst, S.J., Lu, X.: Partial-order-based process mining: a
survey and outlook. Knowledge and Information Systems 65(1), 1–29 (Jan 2023).
https://doi.org/10.1007/s10115-022-01777-3
14. Lepsien, A., Pegoraro, M., Fonger, F., Langhammer, D., Aleknonyt˙e-Resch, M.,
Koschmider, A.: Ranking the Top-K Realizations of Stochastically Known Event
Logs. In: ICPM 2024 Workshops. LNBIP, Springer, Cham (2025)
16 M. Aleknonyt˙e-Resch et al.
15. Pernici, B., Bono, C.A., Piro, L., Del Treste, M., Vecchi, G.: Improving the analysis
of the judiciary performance - the use of data mining techniques to assess the
timeliness of civil trials. Int. J. Public Sect. Manag. 37(1), 59–76 (2024). https:
//doi.org/10.1108/IJPSM-02-2023-0058
16. Pernici, B., Campi, A., Dilettis, M., Gerosa, P.: Why are Italian trials taking
so long? A process mining approach. In: Ital-IA 2023: 3rd National Conference
on Artificial Intelligence, orga- nized by CINI. CEUR-WS.org, Pisa, Italy (2023),
https://ceur-ws.org/Vol-3486/34.pdf
17. Piest, J.P.S., Bemthuis, R.H., Cutinha, J.A., Arachchige, J.J., Bukhsh, F.A.: A
Method for Bottleneck Detection, Prediction, and Recommendation Using Process
Mining Techniques. In: E-Business and Telecommunications, vol. 1795, pp. 118–
136. Springer, Cham (2023). https://doi.org/10.1007/978-3-031-36840- 0_7
18. Reinkemeyer, L. (ed.): Process Mining in Action: Principles, Use Cases and Out-
look. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-40172-6
19. Unger, A.J., Neto, J.F.D.S., Fantinato, M., et al.: Process mining-enabled ju-
rimetrics: analysis of a Brazilian court’s judicial performance in the business
law processing. In: ICAIL ’21. pp. 240–244. ACM, São Paulo, Brazil (2021).
https://doi.org/10.1145/3462757.3466137
20. Vercosa, L., Silva, V., Cruz, J., et al.: Investigation of lawsuit process duration
using machine learning and process mining. Discov Anal 2(1), 9 (2024). https:
//doi.org/10.1007/s44257-024-00015-0
