1 Introduction
1.1 Research challenges
-
Similar to design patterns in software engineering [9], PIPs constitute abstract concepts that may or may not be useful in a particular context. Experience from other scenarios, which may widely differ from the one at hand, is thus not sufficient to take reasonable decisions on the implementation of organizational changes or process-aware information systems (PAISs).
-
Ex-post evidence is usually obtained from persons involved in the respective implementation projects. In turn, this leads to a source of bias. Moreover, a priori assessment allows addressing a far wider spectrum of PIPs. In particular, it is not necessary to complete implementation projects before a PIP can be assessed.
-
Effective PAIS development requires to consider process improvement potentials before any implementation effort is incurred. Accordingly, PAIS development should start with a requirements definition which, in turn, is based on adequate process design considering relevant PIPs.
1.2 Contribution
-
The proposal we presented in [13] provides a standard approach to evaluate the impact of PIPs on organizational objectives for specific application scenarios. Thus, it supports well-founded decisions on the implementation of corresponding adaptations of business processes. Moreover, it contributes to bridge the gap between generic PIPs proposed by the BPM research community and real-world application scenarios. This way, it enhances the practical relevance of proposed PIPs.
-
It considers scientific rigor by applying an appropriate research framework.
-
It reflects practical requirements, which could be demonstrated through an experience report covering a substantial real-world business process.
-
It extends the presentation of the sample case used for our experience report with empirical results based on process mining [11]. Thus, it illustrates the application of this technology to a practical case.
-
It provides a discussion of open challenges regarding process improvement tools.
-
It provides a more profound discussion of the applied process improvement methodology, thus rendering the approach more accessible for application to other scenarios.
-
It discusses the complete set of process improvement measures that resulted from the application of PIPs to the sample case.
-
It describes the results obtained when revisiting our process improvement measures 14 months after having completed the initial process improvement project.
-
It provides a substantially extended discussion of related work.
-
It comprises a more detailed reflection of our results against Challenges 1–3, as well as an assessment of the general applicability of our approach. This includes a discussion of relevant limitations and strategies to address these.
2 Sample case: applications management process
-
The process map is based on events logged in the PAIS. Not all events directly reflect a corresponding activity in the process model, and identifiers of events might differ from the ones of corresponding activities. There may be activities not reflected in a logged event or events not triggered by an activity from the process model.
-
The process map shows the actual frequency of events in the data sample. Thus, it reflects as-is process execution, which may differ from to-be process design as recorded in the process model.
-
The process map needs to be interpreted with the support of experienced stakeholders. In our sample case, for example, application refusal events are used to purge the database of received applications to comply with privacy regulations. Further, not all hirings are handled through the corresponding end events. Issues like these need to be understood when interpreting the process map. However, this understanding is useful for process improvement as well.
3 Methodology
3.1 Problem statement
3.2 Research design
4 Sample case: process improvement patterns assessment
4.1 Organizational objectives
4.2 Process improvement objectives (PIOs)
-
Not approving a job offer after a successful interview may be caused by defective steering of capacities (i.e., job vacancies), defective communication of terms to be offered, or defective review of application documents.
-
Job offers declined by applicants mostly means that the applicant does not approve of conditions offered, did not have a good impression during the application process, or has decided to take another job offer.
PIOs | Rationale |
---|---|
Reduce manual processing effort
| Emerging potentials in terms of reducing process enactment effort per instance should be addressed |
Reduce failed approvals
| Final approval of job offers by senior management fails if there are issues regarding vacancy management, reconciliation of terms, or checking of documents. The probability of these “late defects” should be addressed |
Reduce cycle times
| The probability of applicants’ obtaining and taking alternative job offers increases with time. Therefore, cycle times between applications being received and job offers made should be as short as possible |
-
Cycle time is measured between receipt of the application and the ultimate feedback to the candidate, whereas the withdrawal sample refers to candidates that declined a job offer thereafter. An impact of the occurrence of a withdrawal on cycle time can therefore be ruled out.
-
There is a plausible explanation for longer cycle times causing withdrawals: It is possible that candidates find another job while waiting for feedback after an interview. This explanation is substantiated by data on withdrawal reasons collected for a sample of 51 withdrawals between October 2013 and January 2014 for one business unit. The sample covers only cases where a reason was given for the withdrawal. In 33 out of 51 cases, the reason cited was a job offer by a third party, and we may assume that longer cycle times provide more opportunities for candidates to find alternative employment.
-
We discussed potential additional independent variables with a positive effect on both cycle times and the probability of withdrawal with HR management and obtained no evidence on such variables. HR managers even mentioned that particularly sought-after candidates, who can be expected to quickly obtain alternative job offers, are handled with higher priority by business units. This effect might even “hide” part of the correlation between cycle times and probability of withdrawal. However, quantitative evidence on this issue is not available.
4.3 Process improvement measures (PIMs)
PIOs | Applicable PIMs with comprised PIPs |
---|---|
Reduce manual processing effort
|
PIM 1 (Application management automation): Task automation\(^\mathrm{a}\), routing automation |
Reduce failed approvals
|
PIM 2 (Utilization and capacity management): Empowerment\(\mathrm{a}\), knockout\(^\mathrm{a,b}\)
|
PIM 3 (Standardized terms and conditions): Triage\(^\mathrm{a}\), buffering\(^\mathrm{a}\)
| |
Reduce cycle times
|
PIM 4 (Managing interview feedback cycle times for successful applicants): Control addition\(^\mathrm{a}\), routing automation, escalation procedure |
PIM 5 (Improving application routing): Case manager\(^\mathrm{a}\), knockout\(^\mathrm{a}\), mitigation of repetitive loops |
-
Implementation describes steps required to realize the measure.
-
Business value appraises the expected implications considering the impact on PIOs as well as implementation effort.
-
Stakeholder verification describes the results of validating the PIM through interviews with respective stakeholders.
4.4 Implementation results
4.5 Deployment of tools for empirical process analysis in practice
-
Pattern analysis allows comparing multiple process enactment variants [32] including their actual frequency. For example, with regard to repetitive loops (cf. PIM card 5), this functionality would be very useful to identify and prioritize process variants that should be restricted or eliminated.
-
Compliance rules modeling allows describing relevant regulations for business processes in a way sufficiently formalized to automatically check whether these regulations have been adhered to in a process enactment data sample [22, 33]. As an example, consider the requirement of obtaining approval before job offers are issued.
-
Approximation of manual effort facilitates amending event-based process maps with the underlying manual processing effort. This would tremendously enhance the utility of corresponding analyses and could be achieved by enriching event types with assumptions on the corresponding activities. By matching a material sample log against total capacity used for processing (the so-called baselining in practice), the required degree of validity for the assumptions made could be achieved.
-
Automated regression analysis allows finding correlations between characteristics of data samples (e.g., between the number of contact partners involved and cycle times). If characteristics are derived from PIOs, respective drivers for process improvement can be identified automatically.
-
Sample delineation addresses the issue of restricting a data sample to exclude process instances without particular characteristics, such as the presence of start and end events. Since this topic is important to ensure the validity of analyses, tool developers might want to consider guiding users through the sample delineation procedure by way of an appropriate user interface.
5 Related work
5.1 Validation of process improvement patterns
5.2 Identification of process improvement patterns
5.3 Additional aspects
6 Discussion
6.1 Revisiting research challenges
-
The ex-post perspective seeks to narrow down the set of PIPs to a selection of aspects with demonstrable empirical relevance in a wide variety of application scenarios, thus following common scientific practice.
-
The a priori perspective seeks to accommodate a comprehensive set of PIPs, but limits assessment to one particular application scenario. It thus “constructively validates” only a limited set of PIPs at a time.
-
Is the business process substantially relevant to the organization, e.g., with regard to the output produced or the cost volume incurred?
-
May the organization assume that there are improvement potentials in the process, for example, when considering existing problem reports or benchmarks [30]?
-
Are there particular circumstances that require analyzing the process anyway such as, in our case, intentions to replace the underlying PAIS?
6.2 Relevant limitations
-
On a more detailed level, the business value of PIPs is appraised considering the business process and the scenario addressed. That is, the general assessment approach is refined specifically for each application scenario. Thus, it is not possible to fully replicate the same assessment approach in other settings, which limits the possibility of empirical validation. In other words, the validity of predictions on the business value of a particular PIP in a particular setting cannot provide assurance on the validity of predictions on other PIPs in other settings.
-
Revisiting PIMs after implementation will only allow identifying “false positives,” i.e., PIMs that did not deliver the business value expected. “False negatives,” i.e., PIPs not chosen for implementation which would have delivered a positive business value, will always remain undetected.