Towards improving agility in public administration

Citizens’ expectations with regard to public administration are changing, and technical developments in the context of digitization are leading to growing demands on public administration. This results in high expectations of public administration in terms of citizen communication, participation, and transparency. The increasing demands and the resulting topics and projects usually do not fit into the previous structure of public administration because they are new in the way they arise or in terms of content and move across the cross-section of different areas. These topics cannot be embedded in the actual line structure of public administration, as they have to be dealt with across departments and within several responsibilities and hierarchies (Bartonitz et al., 2018). As a result, the work of employees in public administration is no longer exclusively limited to the administrative tasks assigned to them, but project and teamwork is increasingly coming into focus. Methodological knowledge for project work is often not available and are influenced by corresponding influencing factors. Public administrations are often bureaucratically structured and organized. Due to the fixed allocation of responsibilities, the official hierarchy, administrative regulations, and documentation as the main features of administrative organization, they allow little flexibility and can lead to frustration and demotivation among employees (Bogumil, 2008). Agility can be a crucial improvement factor for these mentioned challenges in team-oriented project work in public administrations. The influencing factors clearly show the need for targeted support in the agile transformation process and demonstrate that instruments originating from pure software development cannot be transferred to this context without appropriate adaptation. Public administration is therefore a suitable context for an initial case study, because a particular importance must be attached to targeted support in the agile transformation process that takes the influencing factors into account.

Some studies report results concerning the measurement of success of agile transformation or for measuring agility. The 14th Annual State of Agile Report (Digital.ai, 2020) shows how success of agile transformations is measured in different types of organizations. The top measures are ranked as follows: customer or user satisfaction, business value, on-time delivery, quality, and business objective achieved.

The literature also mentions some maturity models, which evaluate the agility of an organization and identify approaches for improving it. In 2007, Sidky et al. (2007) published the Agile Adoption Framework, which is based on the principles of the Agile Manifesto (Beck et al., 2001). The framework includes the measurement of agility with the Sidky Agile Measurement Index (SAMI), and a process model for introducing agile practices within an organization. In 2007, Packlick (2007) defines a maturity model—by way of the Agile Maturity Map—consisting of the components’ target areas, goals, acceptance criteria, and maturity levels. Goals and their acceptance criteria are presented in the form of user stories. In 2009, Patel and Ramachandran (2009) developed the agile maturity model. The model defines five maturity levels as well as an administration process based on self-assessment through assessment questions. The maturity levels are used to define goals for the software development process. In order to achieve the goals, the model shows areas of improvement at each maturity level. In 2010, Benefield (2010) created a model that distinguishes agile maturity for seven dimensions: automated regression testing, code quality metrics, automated deployment and backup, automated builds and configuration management best practices, interlocked delivery and interface integration testing, test-driven development (TDD), and performance and scalability testing. Each of these dimensions is assessed on the basis of five levels of maturity. In 2011, Yin et al. (2011) developed the scrum maturity model, which is based on five maturity levels. For each level of maturity, concrete objectives and practices are defined, which have to be implemented by the organization in order to achieve the target. Fontana et al. (2015) coined the Progressive Outcomes Framework in 2015. The framework distinguishes six progressive outcomes: practices, team, deliveries, requirements, product, and customer. For each of these outcomes, maturity levels are defined, varying in number between three and four. In 2007, Qumer et al. (2007) provided a maturity model called Agile Adoption and Improvement Model (AAIM). This is an extension for the Agile Software Solution Framework (ASSF) (Qumer & Henderson-Sellers, 2009). The ASSF describes—on a high level of abstraction—an approach for the creation and adaptation of an agile software development process.

In addition to the aforementioned maturity models, there are also questionnaires that measure agility. In 2009, So and Scholl (2009) published the Perceptive Agile Measurement questionnaire. The questionnaire measures the perceived agility independent of the concrete, technical tools that are used in software development. The items of the questionnaire measure eight factors: iteration planning, iterative development, continuous integration and testing, stand-up meetings, customer access, customer acceptance tests, retrospectives, and co-location. In 2015, Gren et al. (2015) developed a questionnaire that can be used to extend the Agile Adoption Framework (Sidky et al., 2007). The questionnaire replaces the procedure for analyzing organizational readiness, according to Sidky et al. (2007). In contrast to the organizational readiness, the questionnaire is not used to determine the agile potential, but to investigate the current degree of agility.

The aforementioned models for improving and measuring agility have the disadvantage that they do not take into account the context of the particular organization. Therefore, these works are not specialized in the context of public administrations described in the introductory section. Especially in agile process models, context plays an important role. For instance, there are studies that show that the culture of an organization has an impact on the use of agile practices (Gupta et al., 2019). Moreover, some of the models referred to have been developed for software development projects. In the context of public administration, however, software is not developed in every project. Therefore, some of the models cannot be applied to improve agility in public administration without adaptation.

Research regarding the measurement of agility in the context of public administration is rather limited. In 2018, Mergel et al. (2018) conducted a systematic literature review regarding agile government. They identified four application areas of agility in government (agile software development, agile project management, agile acquisition, and agile evaluation). Mergel et al. (2018) state that the category of agile evaluation is largely underexplored in a governmental context. They found two papers which deal with the topic of evaluation. One of these, the work by Dahmardeh and Pourshahabi (2011) is related to this work. They provide a knowledge-based framework for the measurement and assessment of public sector agility and use Kearney’s Model of Agile Government as a basis for creating their scales. In contrast to our work, Dahmardeh and Pourshahabi (2011) view the topic of agility from a different perspective (agile manufacturing) and use a different basis for their scales.

The analysis of related work reveals that there is a need for support in agile transformation. So far, the authors have not been able to find any work that deals with the application of models for measuring or improving agility in the context of public administration that is based on agile values. Considering the related works in more detail, only the work of Dahmardeh and Pourshahabi (2011) from 2011 deals with a measurement of agility in the context of public administration, but this model does not consider agile values. Particularly due to the special features of public administrations mentioned in Sect. 2.1, the main focus should be placed on value orientation in the context of targeted agile transformation. Furthermore, for an application of the questionnaire for measuring agility in public administration, a linguistic adaptation to this context must be made before. Because this context is different to software development, which was taken into account while developing the initial questionnaire. Moreover, due to the diversity of subject areas and tasks in public administrations, teams, and their projects are also of a very different nature. For this reason, it must be possible to adapt the tool to the user-specific context of an individual team within an administration. Known models either do not take agile values into account or they have not been adapted to the context of public administration.

To answer this research question, an appropriate research methodology must be selected. In this regard, the authors chose to perform a case study. The approach to use a case study allows us to collect data in practice and to better understand the context. Complex phenomena with their respective contexts are investigated in case studies (Baxter & Jack, 2008; Goffin et al., 2019; Yin, 2003). According to Baxter and Jack (2008), the research question relates to how-questions and the authors want to cover contextual conditions because the boundaries are not clear between the phenomenon and its context. Furthermore, case studies are not only a valid tool for generating a theory to answer the research question. They can also be used to test and elaborate the constructed theory (Goffin et al., 2019). This work is therefore going to present the findings of a descriptive single-case study embedded multiple units of analysis (Yin, 2003), in which the agile transformation is supported by the questionnaire for measuring agility (Looks et al., 2021). The study comprises three main parts: study planning, data collection, and data analysis. The activities of these parts are summarized in Fig. 1, together with the preliminary work on this issue.

In the following, the authors are going to explain the general context of their case study and then describe the differences between the embedded units (Team A, Team B, and Team C). The interaction between the entities is shown in Fig. 2.

In Germany, the administrations of the regional authorities of the federal government, the federal states and the local governments, which consist of organizational units of different types, are referred to as public administrations (Reichel, 2010). Public administrations in Germany can be divided into planning, law-preparing and law-enforcing organizational units (“authorities”), as well as organizational units with service and production character (“establishments”). As an administrative part of the executive branch, public administration serves to implement and concretize political decisions made by the legislature and is thus accountable to the government. Consequently, the administration is a central component of the authority-sharing organization of the modern constitutional state (Reichel, 2010). The case study was conducted with three teams of a local government. As part of the executive authority, the local government consists of the municipal council and the administrative staff, who are led by democratically legitimized mayors. The differences between the three local government teams that contributed to the case study are explained below.

In the planning phase of this research, it was already clear, based on a literature review and the practical experience of the authors, that the agile transformation in public administration had not yet progressed very far. Therefore, the authors could not select just any public administration for the case study. The authors defined the selection criteria as follows:

The selection of a suitable public administration was made on a voluntary basis. For this purpose, a call for participation was launched via the newsletter of an organization for the establishment of agile methods in German public administrations. This organization consists of an association of practitioners from public administration (municipal, cantonal, and federal administration) as well as from administration-oriented service companies. The authors received some responses to their request for search; however, only one administration fulfilled the criteria above. In this regard, the authors also contacted the public administration, which were examined in the case study. Initial discussions with employees in the studied public administration showed that first experience with agile process models and practices was already available, and moreover, that a step towards an agile transformation had already been taken.

The case study was conducted during the period from 2020/04/27 to 2020/07/16. Table 1 shows an overview of the different methods used to collect data. All studies were conducted remotely, so no face-to-face contact with study participants was necessary.

Agile and Lean paradigms focus on continuous improvement. In the previous work, the authors have investigated how organizations could improve their value delivery by means of applying agile practices to product development (Schön et al., 2015, 2017a, b, 2020). In this context, measurement plays an important role. For this reason, the authors have developed a questionnaire for measuring agility to support agile transformation. The questionnaire for measuring agility measures the current state of agility in software development (Looks et al., 2021). The scales are developed using the agile values as a basis. The questionnaire for measuring agility was formerly developed to support agile transformation in the context of software development as well as digital product development. Applying it to public administrations requires prior alterations, because software is not developed only in this particular context and projects often do not deliver a digital product as a result. In a first step the authors adapted a questionnaire for measuring agility (Looks et al., 2021) to the context of public administration by means of empirical expert validations in several iterations. This activity was conducted in the period from 2020/04/27 to 2020/05/24 (see Table 1). In the first round, 20 experts from different public administrations evaluated the questionnaire’s items in terms of their suitability. The experts judged that 17 of the 28 items had to be adapted because the terms used were specifically connected to software development and thus unsuitable for public administrations. The items considered to be inappropriate were then rephrased by three experts who possess knowledge in the fields of public administration and agile process models. Subsequently, the individual items and the dimensions were reviewed and adjusted once again.

Once the questionnaire had been adapted, it was possible to carry out a measurement with teams in public administration. Initially, two workshops were conducted with study participants to introduce the study and the questionnaire for measuring agility. The authors applied a questionnaire for measuring agility and obtained quantitative data, which was evaluated by means of statistical analysis. The workshops were used to achieve a shared understanding of the dimensions of a consistent weighting of the relevant questions. A shared understanding of the dimensions is important for a measurement with the questionnaire. Only with shared understanding, the respective team can jointly make a weighting and thus set the goal of the measurement. In the following, the items of the questionnaire for measuring agility, which is used to accompany this case study, are presented. Table 2 shows the questions regarding the attitude and experience of the questioned participants. Originally, more demographic questions were planned, but the authors had to shorten them to preserve the anonymity of individual team members in the small sample at the team level (N ≤ 8).

The questionnaire for measuring agility considers the user-specific context. Therefore, the questionnaire contains weighting questions (see Table 3). For each of the six dimensions of the questionnaire, an additional item is defined that queries the importance of the dimension. The weighting questions are rated on a 7-point Likert scale with the following scale points: particularly important, important, rather important, neutral, rather unimportant, unimportant, and particularly unimportant. Weighting questions allow the user to determine which dimension is relevant to his or her context. For example, the product-driven dimension might be less important in an environment where many projects are carried out and fewer products are developed.

In addition to the weighting questions, the questionnaire includes 28 items. With the exception of the communicative dimension, which contains eight items, all dimensions contain four items and form a psychometric scale in the individual dimensions. Those items are assigned to six dimensions (communicative, change-affine, iterative, self-organized, product-driven and improvement-oriented), as shown in Table 4 and are used to assess the agility of a team. Regarding the scale points, a 7-point Likert scale was chosen. For the definition of the dimensions, the agile values were compared with the traditional values. Based on the agile expression of the defined value pairs, the six dimensions for the questionnaire were defined based on a mixed strategy with aspects of intuitive, rational, criteria-oriented and factor-analytical construction (Looks et al., 2021). The individual scale points are verbalized as follows: totally agree, agree, rather agree, neutral, rather disagree, disagree, and totally disagree.

Since the study was conducted in a public administration in Germany, the data were collected in German and translated to English for the purpose of this publication.

Data analysis has been organized as an iterative process in which the authors of this paper have participated. The authors first made the adaptation of the questionnaire for measuring agility in public administration and then started conducting the survey. For this purpose, the authors analyzed the workshop (see Table 1) for the preselection of the weighting questions with regard to the study design and the impact on the embedded units of the case.

Data gathering for the questionnaire for measuring agility was conducted as an online questionnaire. The authors used the tool LimeSurvey to organize the online assessment. The tool helped us collect the data and make it available for analysis purposes. The following statistical formulas were used to analyze the questionnaire: mean, weighting, standard deviation, variance, confidence, and confidence interval (p = 0.05). The data were examined individually for each embedded unit as well as a cross-case analysis. The evaluation was carried out in the individual dimensions (communicative, change-affine, improvement-oriented, self-organized, product-driven, and iterative), due to the partially large variance of the questionnaire results. This enables evaluation at the team level. By comparing the mean of the participants at team level, outliers within the questionnaire can be identified. Outliers are values that deviate from the remaining data. The term outlier is not a well-defined mathematical term, but a subjective observation concerning certain observations made after data collection (Gather, 1980).

Figure 3 shows a comparison of the calculated mean values (rating obtained through the questionnaire) of the dimensions for Team A and the defined weighting values (obtained from the workshops before). The mean value and all statistical results determined for a dimension are calculated by including all participant responses and thus by including the answered data points of a team in the dimension. The error bars represent the calculated 95% confidence interval.

In addition, Table 6 presents the statistical data for Team A. In light of all the dimensions, the improvement-oriented dimension is rated highest by the participants with a mean of 1.85, and the product-driven dimension is rated lowest with a mean of 0.58. However, the product-driven dimension was assigned the scale value “neutral” (numerical value 0) in the weighting, hence overall the mean result of the dimension is better than the defined weighting value. Likewise, this applies to the iterative dimension with a mean of 0.95, which was rated better compared to the weighting “rather unimportant” (numerical value -1) in the questionnaire. The communicative, change-affine, and self-organized dimension were rated significantly lower in terms of weighting.

A closer look at the evaluation reveals an outlier in the communicative, change-affine, improvement-oriented, and product-driven dimension. This data of the outlier is provided by the same participant whose assessment of the items deviates significantly from the assessments of the other participants.

Figure 4 highlights a comparison of the calculated mean values (rating obtained through the questionnaire) of the dimensions for Team B and the defined weighting values (obtained from the workshops before). The mean value and all statistical results determined for a dimension are calculated by including all participant responses and thus by including the answered data points of a team in the dimension. The error bars represent the calculated 95% confidence interval.

Table 7 presents the statistical data for Team B. The improvement-oriented dimension achieved the highest score of all the dimensions, with a mean of 0.94. This dimension was previously weighted by Team B with the scale value “important” (numerical value 2). The lowest mean value was given to the iterative dimension, with a mean of -0.45. However, with regard to the previously determined weighting with the scale value “rather unimportant” (numerical value -1), the iterative dimension was able to exceed its assigned weighting value. The communicative, change-affine, self-organized, and improvement-oriented dimensions remained significantly below the defined weightings, with the mean values determined (see Fig. 4).

The authors also analyzed the data in greater detail in order to identify possible outliers. Again, the authors found an outlier in the communicative, change-affine, improvement-oriented dimension. For Team B, the data of the outlier is provided by the same participant whose assessment of the items deviates significantly from the assessments of the other participants.

Figure 5 shows a comparison of the calculated mean values (rating obtained through the questionnaire) of the dimensions for Team C to the defined weighting values (obtained from the workshops before). The mean value and all statistical results determined for a dimension is calculated over all participant responses and thus over the answered data points of a team in the dimension. The error bars represent the calculated 95% confidence interval.

Moreover, Table 8 presents the statistical data for Team C. The evaluation of the questionnaire results of Team C shows that the improvement-oriented dimension was rated the highest of all dimensions, with a mean of 2.31, and exceeds the previously defined weighting value “important” (numerical value 2). The iterative and the product-driven dimensions show a higher mean compared to the established weighting value. The mean values of the communicative and the change-affine dimensions are significantly lower compared to the corresponding weighting values.

The detailed analysis of the data indicates an outlier in the self-organized dimension. The data from the participant’s assessment of the items deviates significantly from the assessments of the other participants.

The following is a cross-case analysis of the questionnaire results of the three teams under investigation. Figure 6 shows the mean values determined for the three teams in the respective dimensions and the associated 95% confidence intervals. Figure 6 illustrates that, for the communicative dimension, Team A was able to receive the highest mean value in terms of agility. On the other hand, in the change-affine dimension, Team C showed the greatest progress in agile transformation. Moreover, in the iterative dimension, Team B achieved the lowest mean value. Overall, this dimension was weighted with the scale value “rather unimportant” (numerical value − 1). All three teams exceeded this previously defined weighting. The self-organized dimension achieved a mean value > 0 in all three teams. Furthermore, Team C showed good progress in the agile transformation of this dimension. In the product-driven dimension, Team C achieved the highest mean value compared to the other teams. In the improvement-oriented dimension, all three teams showed good progress in agile transformation. Team C was able to exceed the previously defined weighting value of “important.”

The evaluation of Team A reveals a considerable need for action in the change-affine dimension. The mean value determined for this dimension is 0.65 and represents the largest difference with regard to the defined weighting of the dimension with the scale value “particularly important” (numerical value 3) (see Sect. 4.1). Similar to this, Team B shows potential for optimization in the change-affine dimension (see Sect. 4.2). In the Team B, the highest deviation from the weighting value is represented by the change-affine dimension as well, with a mean value of 0.3. This dimension was given the scale value “particularly important” (numerical value 3) and thus indicates the strongest need for action. Team C shows mean values > 1 in all dimensions (see Sect. 4.3). This already indicates a solid level of agility. The weighting values are almost reached in the dimensions. A need for action can be derived concerning the communicative and the change-affine dimension, since the mean value is below the weighting value and a high weighting with the scale value “particularly important” was previously defined. In all three teams, the improvement-oriented dimension was rated highest. This already indicates good progress in the area of improvement orientation. With regard to the assessment of the questionnaire results of the individual items in the improvement-oriented dimension, it is obvious that retrospectives are already being carried out and team members are actively participating in the continuous improvement of the project.

Due to the high variance in the dimensions, a participant-centered evaluation was conducted in the individual dimensions of the teams. The participant-centered evaluation shows whether there are any downward or upward outliers within the team and in the respective dimension. Outliers may indicate problems in the team dynamics. In Team A, one participant can be clearly identified as a downward outlier within the questionnaire. In the communicative, iterative, self-organized, product-driven and improvement-oriented dimension, the result values of this participant deviate significantly from the result values of the other participants. When looking at the questions regarding the attitude and experience of the participants (see Table 2), this participant rates himself as an advanced beginner in terms of his competence in agile digital product development. The participant rates his thinking and acting in his daily work as “not agile.” All other members of Team A, on the other hand, assess their work as “rather agile.” Furthermore, it is obvious that another team member in Team A rated all dimensions highest. This participant rates himself as “competent with regard to agile development of digital products.” Based on these results, Team A shows both upward and downward outliers. This suggests a different perception within the team with regard to the agile way of working and may indicate dysfunctionalities of this team, which should be further investigated. Further investigation of the possible dysfunctionalities can be done, for instance, with the help of moderated group discussions. In these group discussions, the team has the opportunity to critically reflect on its own agility by discussing the measurement results.

The achievement of a low confidence interval can thus show whether there is a shared understanding in the small teams with regard to the state of the agile transformation process. The measurement of agility in a team does not only mean a pure measurement of agility in six dimensions but also serves as a basis for further discussions in the team and thus as a starting point for the conceptual framework.

A conceptual framework was developed to answer this research question (Baxter & Jack, 2008). Figure 7 shows our conceptual framework of how agility can be measured in public administrations.

The left side of Fig. 7 displays the initial situation. A description of the context in the current version can be found in Sect. 3.2. In this case study, three teams were investigated as embedded units. The authors used a questionnaire to measure the degree of agility of these three teams. Even the use of the questionnaire leads the participating teams to an intensive confrontation and reflection of their own working methods with regard to agility. The evaluation of the questionnaire is also associated with the development of measures to improve agility. For this purpose, a contrast is made between the objective (see Table 5) and the measured result (see Sects. 4.1, 4.2, and 4.3). The deviations of these two values give an indication of topics that should be introduced and discussed and require further analysis. Additionally, measures to improve agility are proposed and applied in the context under study. The result is the optimized context (see right side of Fig. 7) which uses hybrid process models for product and service development. Hybrid process models combine different methods and practices and are made up of natural process evolution, which is mainly driven by experience, learning, and pragmatism (Kuhrmann et al., 2017).

The measurement of agility showed that all three embedded units have optimization potential in the change-affine dimension. Therefore, measures are going to be developed collaboratively with the individual teams in further workshops to improve this important aspect of the agile way of working. The process of implementing the measures must be accompanied by an expert (e.g. an agile coach), as it is proven that hierarchical culture has a negative impact on the usage of social agile practices (e.g. facilitate social interaction, collaboration, and direct communication) as well as a negative impact on the usage of technical agile practices (coding/testing-oriented software engineering practices) (Gupta et al., 2019). Furthermore, after applying the measures to improve the change-affine dimension, a new measurement of the agility of each embedded unit is required to evaluate whether the measures show success. Moreover, the questionnaire revealed that there may be dysfunctionalities in Team A. According to Lencioni (2012), it is important to identify and resolve dysfunctionalities in teams in order to prevent team failure. The failure of a team within a public administration does not support the improvement of agility, given that the failure could be used as a negative example against the adoption of agile principles. Thus, it is necessary to create a recommended course of action for dealing with the findings of the application of the questionnaire for measuring agility. All teams should have the opportunity to develop continuously. A prerequisite for this is a cultural change in which the team members and their interactions are prioritized, with the aim of continuously sharing and learning from each other. The authors recommend consulting an agile coach for the implementation of the aforementioned measures to improve the public administration in terms of agility. Agile coaches are skilled in leadership qualities, project management skills, as well as technical skills and have an expertise in agile methods. Agile coaches play a significant role in addressing challenges in an agile transformation, such as resistance to change (Stray et al., 2020). They are responsible for removing barriers to team autonomy in agile teams and make agile meetings more valuable (Stray et al., 2020).

Public administrations are generally not classified as democratic type organizations. Democratic type organizations are more suitable for the use of agile process models (Siakas & Siakas, 2007). Therefore, the adoption of agility in public administration remains a challenge. The authors are aware of the fact that transformational change is a constant and revolving journey. For this reason, the measurement of progress and the achievement of a shared understanding plays an important part in the continuous improvement of an organization. The change from a plan-driven to a value-driven culture (Schön et al., 2017a), in particular, is a difficult obstacle for a hierarchical environment. However, this change is absolutely necessary for a successful agile transformation.

The authors are aware that public administrations belong to a strictly regulated area and that the laws and regulations are highly country-dependent. Therefore, one could argue that the results may not be transferable. With regard to generalizability, the authors can make the following statements. The authors developed their model for measuring agility in public administration inductively based on the aforementioned studies. The adaptation of the questionnaire for measuring agility for the context of public administration was evaluated by means of an expert survey. Therefore, this part of the results is applicable to other cases as well. If the conceptual framework is looked at more closely (see Fig. 7), it becomes clear that the elements can be traced back to the literature. The description of the context of public administration is based on the description of a context in Yin (2003). Furthermore, the three circles (agile working team, measure degree of agility at team level, and apply measures to improve agility) are based on the build-measure-learn feedback loop by Ries (2011), adapted to the improvement of an embedded unit. Both elements can be applied by any embedded unit. To this reason, the authors are confident that the findings are applicable more widely.

In contrast to this, the measurements with the questionnaire are tailored to the three teams studied and are difficult to be transferred to other cases due to the specific context in which the teams work. Only the measures identified are outcomes which are dependent on the case studied. Hence, the identified issues in terms of agility might not be transferable to other teams.