Based on the results of our interview study, this section discusses the important findings under two groups: general findings and cluster specific findings. Then, it provides answers to the research questions. Finally, in Fig. 8, we share the updated structure of the proposed model after an iteration over the discussed findings. General findings touch to the important remarks with respect to the overall structure of the proposed model, whereas cluster specific findings reflect some of the important patterns observed regarding the clusters. Figure 7 is formed by merging the aforementioned tables in their order of presentation. This consolidated view is provided to allow the reader to observe certain vertical patterns that can be associated with expert profiles. The complete table reflecting the fitness of each cluster with respect to the principles they are positioned under, as well as whether they also fit under multiple principles can be found at the URL provided in the Sect. 4.
5.1 General Findings
Both Principle and Cluster Completeness are Highlighted. Although some additional remarks and suggestions were provided by the interviewees, all of the experts appreciated the completeness of the model elements.
Descriptive Texts and Exemplary Aspects Help Defining the Boundaries. Experts provided positive feedback on being given clear cluster descriptions, example practices, artifacts and aspects associated with clusters so that they can easily establish the context of a cluster.
There Is No “One, All Agreed Positioning” of the Clusters. The concepts and practices of the agile software development methodology are perceived very differently based on the background and experience of the individuals.
In Fig. 7, we observe that most of the average cluster values are greater than or equal to 4.00. This is expected as the underlying elements of the model are extracted from the scientific literature. Where the average values fall under 4.00, it can be observed that it is caused by the distance of the larger organizational units to the implementation level concerns of software development. Even though there is no agreement on a single model structure, the experts provided great insights towards improving the proposed model to capture the reality of an organization. These findings show that the initially proposed model structure was too simplistic for reflecting the reality.
5.2 Cluster Specific Findings
Technical Excellence Clusters Act as a Prerequisite for Frequent Delivery. Technical excellence is mostly interpreted as the first step towards making frequent delivery possible, as frequent delivery implies a certain level of automation, and involves making architectural decisions.
Technical Excellence Clusters are Perceived as Relatively Less Important for Higher Level Organizational Units. As technical excellence clusters are reflecting more the implementation level concerns, their importance is perceived to decrease as the scale of the organization increase. The lower importance score of the technical clusters on higher levels were therefore not surprising for us.
Customer Collaboration Clusters Contribute to Planning. Especially for large organizational units, customer collaboration is commented to be very important, and is perceived as an enabler of the delivery planning activities.
Human Centricity Clusters are Well Perceived. In almost all of the interviews, human centricity clusters received positive feedback. It is often commented that, people play a central role almost in any process, and if the aim of a model is to capture the reality with respect to agile, people should never be overlooked.
RQ1: Do the Pillars of Principles Sufficiently Cover the Relevant Aspects of Agility in Practice? The principles are evaluated to be complete in terms of reflecting the world of agile. Only one of the experts stated that communicating the purpose of agile transformation and the role of management should be reflected in this structure more explicitly.
RQ2: Do the Clusters of a Principle Sufficiently Operationalize the Principle? The clusters are evaluated to be sufficient and complete in terms of spanning the principles they are positioned under. Only one of the experts mentioned that the budgeting aspects may be necessary to position under a principle appropriately.
RQ3: Does the Importance of Clusters Differ Considering the Organizational Levels? From Fig. 7, we observe that the importance of clusters differ with respect to the organizational scale. This is an important finding as it can help conducting contextually appropriate assessments, where the scale of the organization is considered as a component of the organizational context.
5.3 Threats to Validity
In this section, we discuss the validity threats and our attempts to ensure a high quality of research by keeping these threats minimal. We are aware of the four validity threats namely Construct Validity, External Validity, Internal Validity and Reliability as defined, and tailored to the software engineering domain by Yin  and Runeson et al.  respectively. However, as our methodology is not a case study research, not all four of these validity threats are covered in this section. Rather, we concentrate on the following two aspects as they are found to be more relevant for our methodology:
Construct Validity reflects how properly the examined concept represents the ideas of the researchers. Therefore, should there be any misunderstandings between the researchers and the interviewed parties with respect to the definitions or concepts that are being discussed, they should be addressed. In our interview study, in order to proactively avoid potential misunderstandings, we provided descriptive texts for each cluster, as well as exemplary practices falling under that particular cluster. This approach allowed us to establish the boundaries and the context of the inspected cluster. As our interview partners positively commented on the cluster descriptions and exemplary aspects, we can assume that we were able to properly tackle this threat to validity.
External Validity refers to the generalizability of the derived results from a research activity. In our research, validation of the consolidated agile principles and agile practices is performed by means of expert interviews. This procedure makes the validation susceptible to converge towards and be specific to the potentially strong personal opinions of experts. In order to overcome this threat, first, we derived the principles and practices from scientific literature. This scientifically grounded approach allowed us to have a safety net, in terms of the further validation of the elements within the model. Moreover, as a second attempt to ensure the external validity of clusters and the structure of the model, we have selected experts from different business organizations. Each of these experts has more than 7 years of experience in the domain of agile software development methodologies, and provide their expertise in the spectrum of domains from consulting to software architecture. To conclude, although it is generally accepted that statistical generalizability should not be expected in empirical studies, we have put strong emphasis on ensuring the external validity of our findings.