Process modeling is mandatory for performing process control activities or process redesign [Batini et al. However, a wide range of process modeling languages emphasize different aspects of processes [Recker et al. What different applications of process modeling languages for PDDQM in organizations can be drawn from the literature.
The broad definition of data quality as suitability for use (e.g. [Madnick et al. 2009]) is difficult to measure [Kahn et al. Classifications of data quality dimensions and a discussion of the most frequently discussed dimensions (namely, accuracy, completeness, consistency, and timeliness) can be found in Batini et al. Furthermore, a comparative description of methods dealing with assessment and improvement of data quality is given [Batini et al.
With the investigation of business process modeling notation (also known as business process modeling notation) (BPMN) [Recker et al. According to RQ1, we investigate the application of process modeling languages for PDDQM in organizations.
PDDQM Methodologies
In one process model, two IPs appear to be used for specific purposes (uploading to a website and used to print mailing labels), while the other IP can apparently be used for multiple purposes, as indicated by the use for "running a business ". [Lee et al. Similarly, another example deals with IPs requested for external or internal consumers [Lee et al. Another reason is that the wrong processes can be identified with wrong data [Lee et al.
However, in only one of these cases are the stages of the PDCA cycle used, while the information quality survey is based on TDQM [Davidson et al. In another example, a customer support data quality (CSDQ) methodology is presented that emphasizes user-centeredness [Keenan and Simmons 2005]. In the third case, TDQM phases are included in the action research cycle [Lee et al.
In the other cases, the methodologies are also adapted [Helfert and von Maur 2001; Mielke 2005; Wang et al. Although TDQM proposes the IMS for process modeling, in only two articles the modeling languages are at least derived from the IMS or IP-MAP [Davidson et al. Similar to the methodologies, the process modeling languages – DFDs as well as PFCs – seem to be adapted depending on the specific situation and needs.
In particular, TDQM is referred to as a general quality concept, without it being implemented as a specific methodology [Helfert and von Maur 2001; Kovac et al. Even when referring to the IMS or IP-MAP (within the TDQM methodology), heterogeneous process modeling languages can be observed [Davidson et al. In the primary studies, the COLDQ methodology and the embedded information chain map (cf. Section 2.3) are not used at all.
The application of process modeling languages will be considered in the next section in detail.
Process Modeling Languages
Used process modeling languages and their representation characteristics Article Process modeling Swim time sequence Data quality. Article Process Modeling Swim Time Sequence Data Quality Language (number of trajectory axis element figure in article). On the other hand, there are several options for integrating data quality without the use of IP-MAPs.
Focusing on the process model representation, we consider data quality checks as modeling elements that determine that some form of data quality check is performed at one or more points in the process. For data quality dimensions, specific quality dimensions are visibly incorporated into the production of the process model. In the models presented, the data quality elements used are mostly data quality checks integrated into flow diagrams or diagrams as process steps [Kovac et al.
In one case, data quality checks are integrated as a swimming lane, as the (intermediate) outcome of the process is regularly agreed upon by two parties [Harkness et al. Data quality checks within IMS or IP-MAP are dedicated model elements that can refer to relevant metadata [Ballou et al. In addition to adapting IMS or IP-MAP, there are further approaches that link data quality metadata to process models, but not necessarily to data quality checks [Helfert and von Maur 2001; Kovac et al.
There are further approaches in which process models are embedded and data quality controls are included. Two other approaches explicitly integrate data quality into single process models without using data quality controls. Without using data quality checks, Mielke [2005] provides quality dimensions and metrics to measure data quality within process models.
On an abstract level, the model provides an overview of the most important data quality dimensions for the main processes and departments.
REPRESENTATIONAL REQUIREMENTS FOR PDDQM
In addition to specific data quality control tasks, defined data quality measures between process deliverables are indicated by dedicated arcs. Numbered notations refer to verbalized data flow processes and are associated with data quality dimensions, according to data quality indicators and measurement points with elements of the data distribution process. 2012] provide a formalized meta model, based on BPMN, for assessing the quality of data related to process model tasks.
The data quality of each subprocess is determined, using weighted key performance indicators to measure data quality, based on the most important dimensions of data quality. Additionally, the overall data quality performance score is calculated from the weighted data quality across all processes. Regarding the specific elements for data quality, Table V already provides quality controls and quality dimensions, partially extended by quality metrics.
Within the IP-MAP, quality dimensions and metrics are not prominently included as elements (see Table VI). Further elements for data representation within process models may be necessary in addition to the IP-MAP elements. Finally, Mielke [2005] provides an element for representing a data transformation process, which would be redundant within IP-MAP as it focuses on data production and does not differentiate between data and control flows. .
Unlike the IP-MAP, the process model does not seem to track the production of a specific IP. Again, the differentiation of data and process flow may be important for the consideration of additional process steps not involved in data processing, enabling a broader application of the IP-MAP. Instead of referring to data flows with the predefined naming of the IP-MAP (raw data (RD), component data (CD), IP), data objects' names can be included within the model.
The inclusion of data elements such as text annotations should be considered in the same way as the data quality dimensions, regardless of whether this information should be included in the model.
DISCUSSION
- Data-Quality Management at the Process and Organizational Level
- Methodologies and Process Modeling Languages
- IP-MAP and Its Benefit
- Integration of Different Process Models
- Integration of Data Quality into Process Models and Languages
- Limitations
Therefore, successful reactive data quality efforts should be used as a basis for the development of proactive methodologies. An important step towards proactive PDDQM is the integration of the data quality perspective in widely used process modeling languages (e.g. [Ofner et al. 2012]). Such enhanced process modeling languages can raise awareness and provide a foundation for addressing data quality during process design.
Finally, data quality research identifies generic root causes of data quality problems [Eppler 2001; Lee 2006; Liu and Chi 2002;]. Using general principles instead of the many data quality criteria can reduce the complexity of data quality management. Regardless of whether data quality is integrated marginally or in a holistic way, both seem to be possible with the improvement of existing models or process modeling languages (e.g.
Other modeling elements, such as the different types of gateways, can have an impact on process modeling from a data quality perspective. Based on the identified primary studies, we focus on the integration of data quality in process model instantiations. A visible integration of data quality into process models facilitates the awareness and understanding of data quality issues in cross-processes.
IP-MAP allows transparent integration, facilitating communication between stakeholders (modeling and non-modeling experts) while providing a data quality metamodel for a sophisticated definition of data quality metrics [Shankaranarayanan et al. If integrated into a specific process modeling language, the metamodel must define the possible implementation of data quality control within process model instances. Their process-driven methodology supports linking data quality issues to root causes and determining improvement activities.
The integration of quality data into widely used process modeling languages is at an early but promising stage and would enable a proactive approach to PDDQM from process design.
CONCLUSION
Implications for Practice and Research
Existing methods and process modeling languages can be incorporated to increase familiarity rather than switching to new and unfamiliar ones. Identifying or developing an adequate process modeling language for PDDQM. The integration of different process models is an important and seemingly underestimated issue. Since IP-MAPs are still under development and other process modeling languages are used for PDDQM, the integration of IP-MAPs and other widely used process modeling languages should be addressed (e.g., [ Cappiello et al. 2013; Lee et al. 2007a; Ofner et al. 2012]).
Regarding the future application of current process modeling languages with a focus on data quality, for example, BPMN [Cappiello et al. The Bunge-Wand-Weber representation model allows comparison of IP-MAPs with other process modeling languages [Rosemann et al. To cover the needs of practitioners and to further support the adoption of a new or improved process modeling language, the identified requirements (Figure 3) are a basis for expanded use and further research.
However, to improve the automation, control and support of tools of process modeling languages (eg, [Fahland et al. 2011;. Integration of elements in instances of process models and development of process modeling languages that support PDDQM are both feasible approaches and should be applied according to the purpose of modeling. Moreover, comprehensibility and complexity should be ensured in process modeling [Reijers and Mendling 2011], especially if process models are improved, for example, with data quality aspects [Glowalla and Sunyaev 2013b].
In addition to the process modeling language itself [Recker 2010b], familiarity with existing process modeling languages influences its use [Recker 2010a]. Research into model comprehensibility or complexity, respectively, must assess the right balance between familiarity and usability of a new or improved process modeling language and its instantiations. Several factors must be considered, which could be categorized as contextual factors [Rosemann et al.
Both approaches – improving an existing or developing a new process modeling language – call for additional, differentiated research.
Contribution and Impact