Understanding the importance of collaborative problem formulation for data analytics: Key lessons about leadership,governance,and technology

2.1 Data analytics in the public sector and the importance of collaborative efforts

Data analytics has been explored in multiple ways and contexts. It is commonly referred to as a group of technologies or a framework for intelligence practices that include statistical analysis, data mining, and machine learning methods (Chen et al., 2012; Gil-Garcia et al., 2018; Iqbal et al., 2020). Those practices can help analyze data in more elaborated ways and help organizations extract further helpful insights (Hagen et al., 2019; Chaudhari et al., 2019). The practice is also centered on the notion of “actionable data” (Chong et al., 2018; Bibri, 2019), which are data capable of being really transformative as an enabler for organizational practices, especially for decision-making processes. This notion is not novel and has appeared quite extensively through different terms and narratives, in the knowledge management literature for the past two decades (Davenport et al., 2001; Pee & Kankanhalli, 2016). Analytics has also been associated with several organizational practices and contexts, including marketing (Deng et al., 2015), human resources (Marler et al., 2017), finance (Cokins, 2014), operations (Dubey et al., 2019) and a variety of related disciplines and subdisciplines. The increased interest in analytics appears to stem from recent advances in data processing technologies, which have been able to define new horizons in terms of problem-solving capabilities (Puron-Cid et al., 2016). Incorporating data-driven approaches into the routine has thus become a modus operandi for the modern organization across multiple sectors, industries and disciplines.

In existing public sector research, data analytics appears to have many definitions. Besides being also referred simply as analytics, frequently in the context of big data, there are few specific situations where adaptations to contextual terminologies such as “policy informatics” (Dawes & Helbig, 2015) or “policy analytics” (Höchtl et al., 2016) also occur. Quite often in those contexts, the technical analytic solutions do not appear to be the most important element, but rather the governance challenges that emerge from their use for public policy and public management issues. Consequently, the use of data and information in public administration research appears to follow its own paradigm, often directing the discussion more toward data governance (Wang et al., 2018) and the socio-technical complexities that emerge from that governance than to the technologies that are involved (Fountain, 2001; Andersen & Dawes, 1991; Wang et al., 2018). Among the complexities inherent in the public sector, inter-organizational information sharing and collaboration (Dawes et al., 2009; Gil-Garcia & Sayogo, 2016) and interoperability (Henning, 2018) remain two of the most persisting challenges, that are yet to be deeply discussed in the context of the relatively more recent data analytics research agenda.

Attention to data analytics in the public sector is not uniform across different levels of government, i.e., local, state, and federal. At the local level, attention has been given to its promise in enabling the achievement of smart city goals (Cronemberger & Gil-Garcia, 2019; Iqbal et al., 2020). The promises and hopes towards analytics go beyond the perception of it as a technological novelty; rather recent research has presented caveats such as the risks of wrongfully embracing it as a simple solution for complex problems (Mergel et al., 2016) or treating it as an aspect of information use that is not central to public administration practices, yet still a potentially relevant tool (Hagen et al., 2019)

In the light of persisting challenges such as data availability (Welch et al., 2016), research on promising uses of data has focused on making it more accessible through open data initiatives (Zuiderwijk et al., 2015) and on governance challenges, particularly those associated with larger or growing repositories (Janssen et al., 2017). In this context, the technical infrastructure for data collection, management and governance has also been studied, including the growing interest in IoT frameworks to make data collection more efficient (Jararweh et al., 2020) and the possibilities of applying emerging technologies such as blockchain (Fan et al., 2020). The studies have also explored technological improvements that make cities smart (Sun et al., 2016) through the development of capabilities (Lam & Ma, 2019) and drivers or forces (Angelidou, 2015). In data analytics research, those capabilities may refer to both “human” and “intangible” resources (Gupta & George, 2016), including concerns with finding the right talent to correctly use data (Ekbia et al., 2015) and assessing the relationships of data analytics with factors such as “organizational culture” and “top management commitment” (Wamba et al., 2017).

Those focuses may be especially necessary in a context where multiple stakeholders may need to collaborate to understand public issues and design a solution (Tucker et al., 2017). While not directly referred to as data analytics in local government literature, previous research suggests that collaborative efforts appear to play a crucial role in helping public leadership and citizens to work with data (Susha et al., 2017; Ruijer & Meijer, 2020; Meijer & Potjer, 2018) or engaging stakeholders in sharing data for the public good (Susha et al., 2019). While most previous research has focused on developing a vision for data analytics use in smart city development contexts or on identifying factors that can be used to build that vision, few studies have considered that different cities may be at different stages of development in their data analytics agenda (Harrison et al., 2018). Different organizational realities, as different local governments across the world are likely to represent, may require specific assessments on existing capabilities. As they do evolve and develop towards becoming more data-driven at their own pace, their ability to leverage data analytics may then follow (Janssen et al., 2017). However, empirical evidence on how local governments collaboratively formulate problems for the use of data analytics is still limited.

Existing research can then be categorized into three streams. The first stream is focused on applied studies, dedicated to techniques and technological applications of analytical technologies to problems that local governments face. Examples include the study of data science practices and algorithms (Engin & Treleaven, 2019), artificial intelligence frameworks (Criado & Gil-Garcia, 2019) and applications that help governments understand a variety of challenges such as urban mobility (Wahid et al., 2018) or healthcare systems bottlenecks (Anisetti et al., 2018). The second stream is more theoretical and focuses on understanding how analytics, as an emerging socio-technical practice, influences both management and governance in organizations. It also examines to what extent the effects produced are in accordance with the goals stipulated by government leadership considering public expectations and existing data handling routines. In this research realm, aspects such as smart governance (Barns, 2018), public value generation goals (Yu et al., 2019) and data management practices (Bibri, 2019) are some of the most covered. A third stream of research gives centrality to collaborative efforts (Broccardo et al., 2019) and its arrangements as an organizational practice. Research in this realm is highly interdisciplinary, relying on a multidimensional view of collaboration, its drivers and forces. According to Bryson et al. (2015), for example, there is an interplay between collaboration processes and collaboration structures that is mediated by leadership, governance, capacity and competencies, and technology. These dynamics appears to be affected by political and power conditions, as well as influenced by strategic contingencies, yet not actually determined by them, suggesting that DA practices could be affected by both socials and technical variables, which is consistent with previous research about the use of information technologies (Orlikowski, 2000).

While these three streams are committed to providing practical and theoretical instruments for the study of data analytics and commonly will overlap in the way they conceptualize data-oriented practices, they do not often provide empirical evidence on the practitioners’ experiences and specific collaborative mechanisms that are leveraged when formulating problems to be potentially solved using DA. The current study could thus be situated at the intersection of the second and the third streams, where collaboration practices and the use of data analytics are expected to be affected by multiple social and technical variables. However, our focus here is on the problem formulation process in local governments. By focusing on the problem formulation and the collaborative practices around it, this paper has an intentionally narrower scope and attempts to specifically contribute to the understanding collaborative problem formulation for data analytics use in local governments.

2.2 A socio-technical view of data analytics as a collaborative process in local governments

This subsection argues that the data life cycle can be used to frame data analytics as a collaborative process from a socio-technical view. It focuses on the use of data and the concepts that are related to the different stages in the data life cycle. Due to the variety of applications and contexts in which data can be used, the study of data use spans experiences in business, public administration and a variety of disciplines where certain particularities can be observed. The interdisciplinarity of the topic makes it more challenging to study, but also enriches our current understanding and expands the potential for further research. Data life cycle frameworks are commonly used to comprehensively study practices and processes where data are used. Those frameworks help to design goals, define stages for data management, and also map expected outcomes from data use. Research has approached the topic in two ways. The first approach encompasses comprehensive research, mostly dedicated to refining established models and enhancing the explanatory power of existing theories in the light of new empirical evidence. That is accomplished, for example, through proposing new frameworks that provide a more holistic view of data uses (Ofner et al., 2013). The second approach involves the scrutiny of specific stages of the data life cycle, where research explores and expands the current view on particular concepts and relationships such as data quality (Otto et al., 2012), data collection (Ku & Gil-Garcia, 2018), or the dynamics of data sharing (Cronemberger & Gil-Garcia, 2019).

Overall, existing frameworks outline four similar stages for data use as well as their definitions. Those stages are generally referred to as 1) data collection, generation or creation (Poe et al., 1997); 2) data cleaning and curation (Jagadish et al., 2014); 3) data analysis (Wang et al., 2018) and 4) data management (Babar et al., 2019). As explained before, each of these stages is explored in research to varying levels of depth and examined in specific contexts from distinctive theoretical angles. Such endeavors suggest that, given the multitude of constructs and meanings present in this dynamic and still evolving research domain, examinations of data analytics practices and processes are typically fragmented or not explicitly made. Therefore, we argue that studying DA in the context of the data life cycle in the public sector could help to provide a needed comprehensive perspective. Much is yet to be learned from the different stages and experiences, particularly those in which data use is directly linked or referred to data analytics practices and technologies. Ku and Gil-Garcia (2018), for instance, found that data collection is a critical part of data analytics practices in local governments and argued that only when the necessary and adequate data are collected, can the analysis become feasible and more useful. Hopes of such improved data collection are often associated with technological advancements, including the Internet of Things. However, as expected, not every local government has fully explored these emergent technologies. In addition, according to Gupta and Panagiotopoulos (2019), assessments on operational capabilities, as well as on the extent to which a culture of data collection and use is in place, are still key.

Therefore, we argue that the concepts related to the data life cycle could be very useful to more comprehensively describe data analytics use in local governments. The specific activities, tools, and routines are expected to greatly vary according to the specific characteristics of each local context. For example, some cities may be efficient at collecting data through legacy systems, but may not have structured data management practices in place (Poltie et al., 2020). Others may present solid frameworks for data handling and processing with limited resources, but, in face of increasing external pressure to achieve smart city goals (De Guimarães et al., 2020), they may face capacity issues and struggle to collect data for specific tasks or problems. Being aware of those differences suggests that the capabilities may differ (Pelton & Singh, 2019), thereby requiring different governance arrangements (Bergan et al., 2020) and reveal the need for better adherence to context-specific techniques (Iqbal et al., 2020).

The data life cycle approach helps put the local government experiences with data at different stages in perspective and identify important differences and similarities. In addition, that approach could help to adopt a process view of data analytics, which is adequate to position different local government realities in a broader context. Research on data analytics has focused on both the relevance of data analytics to achieve smart governments and the factors that may influence the success of those data-driven initiatives (Cronemberger & Gil-Garcia, 2019). Across different articles, the general view is that, prior to successful execution of analytics at the technical level, when all the data is in place and ready to be used, organizational factors, such as leadership support and governance efforts to collect data, are critical (Chatfield & Reddick, 2017). Mergel et al. (2016) argue, for instance, that analytical capability development in the public sector can be linked to three goals: a) managing and processing large amounts of unstructured, semi-structured, and structured data; b) analyzing those data into meaningful insights for public operations, and c) interpreting those data in ways that support evidence-based decision making. According to these authors, those capabilities are especially needed in “small jurisdictions” (Mergel et al., 2016). Finally, as mentioned before, collaboration is the key to data analytics success and, therefore, data analytics should be seen as a collaborative process facilitated by leadership, governance, and technology (Bryson et al., 2015; Bowker & Star, 2000).

Based on this review of existing literature, we propose a socio-technical framework that considers data analytics as a collaborative process and a transformational practice, particularly in local governments (see Fig. 1). The proposed framework highlights social and organizational aspects, but also includes those variables related to technical aspects, such as the technology used for data collection, analysis, and representation. Technology infrastructure is not explicitly included but is assumed to be the foundation for the use of DA-related software and applications. Using this view, data analytics practices involve three stages or processes within the data life cycle, namely, (1) data collection and preparation, as the input side of DA; (2) data analysis, and (3) data representations and visualizations, as the output of DA. These processes are practical and operational, and they are the main focus of the first research stream mentioned earlier. As part of the data analytics processes and the outputs that enable decision-making, they are critical to understanding data analytics in a comprehensive manner, but only will happen after the collaborative practices for problem formulation, the main focus of the empirical part of this study as shown in Fig. 1.

OPEN IN VIEWER

When discussing DA practices as a collaborative process, existing research suggests that these practices are likely influenced by three key constructs that are more generally related to collaboration: Leadership, governance and technology. Collaboration is a foundational practice and plays a central role in enabling inter-organizational knowledge and information sharing, but also DA. Leadership is known to foster the strategic use of information in the public sector. Governance is considered to play a central role in orchestrating socio-technical efforts in terms of rules and arrangements that can maximize benefits from the strategic use of information. Finally, technology acts as both an infrastructure, enabling information exchange at the organizational level, and as an artifact that supports the collaborative effort and, in this case, also facilitates the data analytics process within a local government. Each of these variables are related to collaboration (leadership, governance, and technology) and are expected to influence all stages of the data life cycle in the context of DA in local governments.

Local government realities, however, are likely to vary greatly because specific jurisdictions confront policy issues that are relatively unique. Equally specific are the contexts in which problems are examined and formulated, indeed often a dynamic and complex endeavor that involves public leadership and community involvement (Kitchin et al., 2017; van Zoonen, 2020) and sets out to ultimately establish public value (Cronemberger & Gil-Garcia, 2019). While it is self-evident that problem formulation is a critical first step, the evidence backing up this important stage is still limited and domain-specific (Passi & Barocas, 2019). In particular, collaboration often appears as the critical sustaining factor for living labs (Gascó, 2017) and co-producing processes (Grommé & Ruppert, 2021), while socio-technical aspects of collaborative practices in the early stages are not explored empirically very often in local government research. Similarly, data life cycle frameworks focus on operational aspects of data use and tend to overlook problem formulation and the contextual aspects that may influence that phase, clearly an agenda explored mostly by the systems and simulation modeling literature (Black, 2013; Cronemberger et al., 2017). Acknowledging the foundational role of these data life cycle frameworks and the emerging and dynamic nature of policy issues faced in small jurisdictions, this paper contributes by positioning collaborative problem formulation in local governments as significantly important for effective data analytics efforts.

3.1 Data collection and analysis

First, interviews with the stakeholders involved in data analytics were conducted. They were the central source of information for the analysis, contextualizing the case and its specific efforts. The profile of respondents included analysts, policymakers, champions, and local government leaders. The questions followed semi-structured procedures, asking interviewees about the importance of specific factors related to a digital government framework (Gil-Garcia & Pardo, 2005). For example, the respondents were asked to elaborate on points that they considered relevant to data analytics use in the scope of data and information attributes, technological capabilities, organizational and institutional environments, and some external factors, such as citizen participation and political forces.

In addition, a session with community leaders involved in code enforcement for housing provided a topic-centered example of how City Hall goes about approaching problems of public interest analytically and with a participatory approach. During that session, attention was given to the discussion, to the interactions and iterations, as well as to how the topic was being approached. The structure of the meeting and the way different people involved in data analytics discussed each issue and formulated the problem was also observed. Overall perceptions were supplemented with document analysis of openly available electronic sources, such as the What Works City website (What Works Cities, 2017) and specific, information about the i-Team (Innovate Syracuse, 2017). The concepts presented in the literature review section were used to guide the exploration of the case. The data analysis considered pre-established categories based on the literature, particularly the proposed concepts. However, new concepts and relationships that emerged from the qualitative data were also considered, as they pertained to the particularities of local governments as the research setting for this study. Specifically, we found that collaborative problem formulation was fundamental for the use of data analytics in local governments

3.2 Brief description of the case: Data analytics in Syracuse, New York

Syracuse, New York, engaged in a nationwide program called Bloomberg’s What Works Cities initiative (WWC). Through that program, the city became part of a network of 100 municipalities that committed to “enhance their use of data and evidence to improve services, inform local decision-making, and engage residents” (What Works Cities, 2017). Under the motto of “Build a government that residents can count on” (What Works Cities, 2017), this endeavor focused on fostering “best practices” across selected cities, “helping local leaders identify and invest in ‘what works” (What Works Cities, 2017). The selected cities are granted a formal certification and receive funding and support to foster data-driven initiatives from practitioners and researchers affiliated with such centers as the Government Performance Lab at the Harvard Kennedy School of Government and organizations, such as the Sunlight Foundation, Results for America, Govex (John Hopkins University), and the Behavioral Insights Team.

The WWC program uses a framework that is grounded in four pillars: 1) Commit with goals; 2) Measure; 3) Take Stock; and 4) Act. Under each pillar, several steps to achieve key goals are outlined. Participants are expected to embrace the WWC mission and comply with those standards. Under those guidelines, DA sponsors and champions from each city reconvene at a yearly summit where practices and experiences are shared and discussed. Each city has their individual goals and priorities. For instance, whereas Hayward, CA has a “focus on improving safety and quality of life by curbing illegal dumping,” Athens, GA is interested in measuring and communicating “progress on economic prosperity goals”, and Gainesville, FL wants to work “toward transportation and business life cycle improvements.” Syracuse’s involvement with WWC aimed to “improve open data practices and establish and improve performance management programs to improve results for residents” (What Works Cities, 2017). The mayor emphasized the need to “break down data silos that exist in different city departments” and identify when performance “outcomes are not being met”. As information produced by data analytics initiatives comes to life, taking steps to put it into use includes not only benefiting one specific department, but also ensuring that multiple departments can learn from what others are doing as they start tackling their own data challenges. As the city progresses toward that view, Syracuse has positioned itself as a flagship local government in New York, receiving, in 2019, a $500,000 grant to “buy and then replace all of the city’s streetlights with energy efficient light-emitting diodes”, a technology that is known to reduce greenhouse emissions.

In 2015, the City of Syracuse established the Innovation Team (i-Team). Under the motto “changes call for innovation, and innovation leads to progress,” the Innovation Team focuses on initiatives to address public infrastructure problems and foster local economic development. As observed by Bousquet (2017), “the i-Team has a 360-degree view of city projects, serving as a central hub for coordinating initiatives to produce depth, efficiency, and better outcomes for residents”. Initiatives include the installation of street quality electronic devices that address infrastructure issues, such as street potholes, and the development of an early detection system for water infrastructure problems. Often, those problems are interrelated. For example, whenever roads in bad conditions are found, the Department of Public Works may use that information to assess the quality of the water infrastructure underneath (Syracuse i-Team: Upgrading Infrastruc …). The city has also piloted the use of sensors to collect data on potholes; work that was done manually in the past. These goals and efforts have opened a window of opportunities for using data to frame and understand problems that are brought to City Hall, a process that encompasses either further exploration of available data sources or the pursuit of new sources.

OPEN IN VIEWER

The i-Team is structured into three divisions: (1) the Data division, dedicated to working with departments to understand different ways of working with data to support decision-making; (2) the Innovation division, which focuses on partnering with departments to understand operational challenges and design solutions; and (3) the Accountability division, which monitors the work being performed by the data-driven programs and policies. The i-Team also created a publicly available blog to share information about current and past projects. In one segment of a post, the Innovation Team acknowledged that the City of Syracuse “will need much more than to simply buy technology” to become a smart city (Innovate Syracuse, 2017). These goals have been outlined as 1) helping departments to better understand and use data to drive decision-making; 2) think differently and creatively to innovate around specific challenges; and 3) ensure the ongoing advancement of programs and policies (Innovate Syracuse, 2017). Governance efforts to coordinate the goals across those divisions and the different departments they support were still in their early stages, with several respondents acknowledging the ambiguity of preliminary steps and the need for more sense-making. Under a section entitled “Challenges,” the Innovation Team website mentions specific points they consider very important as they move forward with their initiatives, as well as the questions they considered to be relevant to the work they do. These questions include: “Does the community understand what data are being collected and how they will be used?”, “Do people support this data collection?”, “Do our operating departments understand the outputs of data analytics and are they able to make decisions based on them?” and “Does the data being collected help solve a problem we are currently facing?” These questions are asked to clarify the path for data use ahead, and they also help the involved agencies and the i-Team to better understand and collaboratively formulate problems and prioritize both tasks and resources to manage competing demands being put in place by an ever-growing list of public challenges (see Fig. 2).

4.1 Collaborative practices in data analytics: The role of leadership, governance and technology

The interviewees mentioned the role of leadership as having two fundamental purposes. First, at a more macro, political level, leadership support helped kick the WWC initiative off, sending a strong message to the public and the public servants on the extent to which the city government could prioritize data-oriented approaches to policymaking. Most interviewees indicated that the very existence of the i-Team depended on that focus. Secondly, leadership was understood as a proxy of being “self-driven” and “entrepreneurial” in their relation to data analytics. Given the often-stated limitation of information resources and the difficulties in having good access to them, the respondents seemed to hold to a high level of independence and accountability in the projects they were already conducting. The Interviewees revealed that they enjoyed the freedom to go after the data they needed by visiting communities, holding interview sessions, and thinking about innovative ways of addressing stated problems. While many of the interviewees confirmed that such a precise level of engagement was encouraged from the top of the organization, a few respondents did observe that a collective sense of involvement with the problems the team was facing was critical to move things forward positively.

Collaboration was often mentioned as a crucial element for effective data analytics work in City Hall. The respondents stated that the nature of their work is indeed collaborative and relies on participation of both the analysts and stakeholders who are involved or impacted by the issue being addressed by using precise data. By having people collaborate around that data, new understandings of existing data and of the problem itself then could emerge. While mentioned as being very key to the effort, such collaboration was not referred to as a practice that always comes naturally. Rather, it is a result of clearly concerted efforts to bring people to the same room and get them involved. That effort was not without its hurdles because the people and the data needed to address a particular problem were often scattered across different organizations and had to be mobilized in one venue. Therefore, that participation and mobilization required leadership and resources, a condition that was believed to be at least partially mitigated by the jump start provided by the WWC initiative.

Finally, it is worth noting that governance did not emerge or was not cited as being crucial to the people who were directly involved in data analytics. However, two elements that were observed by respondents could be indirectly related to that task. First, the need to have better institutional mechanisms that can bring people together and make data available for use can make data analytics more effective. Details on what those mechanisms could be, however, were not mentioned. Second, coordination emerged as being an important factor. The respondents observed that the unstructured nature of their work demanded some level of self-coordination and a go-get attitude toward the data and its relevant information for a subsequent effective analysis.

4.2 Data analytics and the data life cycle

This subsection presents the results of our case with a focus on the different stages of the data life cycle and how the activities in each stage were important for the data analytics efforts in Syracuse, New York. It is important to emphasize that the activities in each stage of the data life cycle all contributed in one way or another to a better way to use data for decision-making in the city. Also, it was learned that the data life cycle was a useful lens to better understand data analytics needs, actions, and results.

4.3 Problem formulation in data analytics practices

Evidence appears to suggest that problem formulation takes place in multiple ways and as a result of different efforts from stakeholders involved in data analytics use. Those efforts include steps related to data management and data analytics practices, but also organizational forces and context-specific demands related to the problem and the data available or still needed to tackle the problem. Therefore, it seems clear that problems are emergent issues and the result of a collaborative approach to data analytics, rather than simply a premise. Two sub-dimensions that appear to affect clear problem-framing are detailed below.

5.1 Leadership, governance, and technology as enablers of collaborative efforts in data analytics

Evidence of the importance of leadership across City Hall’s DA efforts is consistent with previous research that established the connection between leadership and effective use of information (Cronemberger et al., 2017; Gil-Garcia & Sayogo, 2016). This time, such connection was considered in the light of data analytics as an unstructured number of practices sponsored externally by the WWC initiative and internally by a political and organizational buy-in. This finding also echoes the literature on the importance of leadership when addressing information needs (Cronemberger et al., 2017) and may raise important new questions for how to nurture those forces better and also keep them active. Levels of leadership support could also be assessed whenever cities attempt to achieve specific results by effectively using data analytics.

As part of positive leadership roles and governance concerns, the findings of this study suggest that establishing a data analytics agenda is central to getting stakeholders and resources both involved in specific data problems. In that regard, WWC seemingly contributed as a critical first step toward successfully institutionalizing DA as a formal policy tool in Syracuse (Hagen et al., 2019), indeed one that both symbolically and in practical terms is capable of creating momentum and fostering intra-organizational collaboration to address city challenges. Similar to triggering efforts, such as open government and open data initiatives (Puron-Cid et al., 2016; Ruijer & Meijer, 2020), these findings imply that data analytics and related topics, such as data management (Harrison et al., 2019; Harrison et al., 2018), are central and needed for public leaders who are willing to engage with DA and want to successfully push a smart city agenda (Poltie et al., 2020; Sun et al., 2016).

As far as these leaders’ technical experience with DA use, one important aspect relates to the qualitative nature of their DA work. Interviewing and coding of unstructured data were a common and central aspects of the conceptualization and formulation of problems. That was evident through many Post-it notes used for group dynamics and “human-centered design approaches” that were found in the rooms where policy meetings were held (see Fig. 2). Data analytics in Syracuse appeared to be highly interdisciplinary, human-centered, and focused on diverse information artifacts, including spreadsheets, maps, and drawings, as well as formal and informal conversations that helped guide the analysis. Much of the data collected informed overall policy-design as well as specific processes through which more data could be collected. That focus could be observed in a session with code enforcement personnel, which was held to not only revisit and reframe residential problems in the city, but also to identify what information could be missing for better analysis of their data collection processes. Following a debate-mediated structure, the i-Team attentively listened to concerns and opportunities for improvement and took notes on perceptions. These notes were later used to define what the next steps for data collection could or should be.

Additional evidence that suggests the foundational importance of problem formulation in data analytics practices comes from the fact that the interviewees were very clear about the scarcity of necessary information to do their analytical jobs. For many of them, that meant that parts of the problem to be solved were not fully understood and needed further investigation. They attempted to better conceptualize the problem and characterized the process as more of a “learning-on-the-go”, a strategy that involved interviewing citizens and organizing information obtained from workshops run in the city hall to obtain further information about a particular problem. Finally, information scarcity also created incentives for in-person visits to partnering agencies, where key staff members would potentially have important data sources or have access to that data in legacy IT systems. According to some respondents, the opportunity to understand information generation processes also contributed to add context to available sources and, in doing so, facilitated problem formulation. Data analytics outputs would further benefit from this effort because a more informed analysis could lead to richer perspectives even when having limited data.

5.2 Data collection and representation as structuring collaboration practices

The results suggest that data management in data analytics does not necessarily follow a structured process. Rather, it is a result of dynamic practices that unfold and adapt over time. That development occurs as data analytics practitioners explore existing and new sources of data. With that exploration can come a greater level of understanding of the problem being solved. It has been suggested that data analytics in small jurisdictions is participatory, closer to co-creation (Yu et al., 2019) and innovation practices observed in smart cities research rather than mostly technological as explored in literature on analytics from a technical perspective. From a public policy design standpoint, data management processes in data analytics use in local government may need to be more reliant on the ability to deal with unstructured data, than with assuming that all the data needed to solve a particular problem will be found in a single place and be ready to be used for a clear purpose. The scattered nature of information resources echoes research on the role of information sharing and integration to successful digital government projects (Dawes et al., 2009; Gil-Garcia & Sayogo, 2016).

It is also important to highlight the centrality of data in collaborative processes (Susha et al., 2017). Although participation seemed to be a key element, information produced and manipulated by data analytics practitioners were the core output and, as such, that information needed to be systematically documented. Such documentation was reflected in the number of Post-it notes used during data collection interactions and the subsequent documents created from the information gathered in those sessions. The literature refers to those artifacts as “boundary objects” (Black, 2013; Cronemberger et al., 2017), but the research discussing this information science concept in the context of data analytics is still scarce. Finally, the more qualitative approach portrayed by Syracuse’s experience also appears to contrast with research that has been more focused on technical and quantitative capabilities than on the ability to create important information from soft, non-digital data. That scenario could be the case for unstructured innovative practices, such as co-creation frameworks for smart cities (Yu et al., 2019), but still always considering the specificities of data analytics in local government contexts.

5.3 Adding collaborative problem formulation to data analytics practices in local governments

Syracuse’s experience is enlightening for several reasons. First, the evidence suggests that DA is not centered at technological artifacts only, but also occurs at collaborative practices through which raw data and information on public problems are shared and used. In that process, the formulation of problems helps define the guiding steps for data collection and, subsequently, for data analytics outputs. Such a process is markedly iterative, with multiple sessions dedicated to knowledge and information sharing, which then leads to a more complete understanding of specific issues. There is also evidence that collaborative problem formulation is a central practice and could be considered as a foundational step of DA as a process. Such efforts are normally qualitative in nature, relying on multiple iterations through unstructured data, and they facilitate collective sense-making and alignment with regards to defining the exact problem that needs to be addressed. This finding is consistent with previous research that acknowledges the importance of problem formulation and how the process can affect the overall results of data analytics (Passi & Barocas, 2019).

As per the findings, the problem formulation phase of DA still needs examination at different levels, including the role of political leadership and the public whenever setting priorities for data analytics use. Externally oriented initiatives, such as WWC, seem to have influence and offer political endorsement at the local level. Internally, a focus on problem formulation and data needs (Harrison et al., 2019) seems to prevail over technology, but it also seems to rely on important soft skills and specific organizational characteristics, including the ability to bring different stakeholders to collaborate around data and produce innovative and useful information. A theoretical view that positions DA as a socio-technical process and a transformational practice may lead to a variety of workable and useful models and frameworks. For instance, a comprehensive framework of DA use based on the Syracuse experience should be attentive to such concepts as collaborative practices, including the role of leadership, governance, and technology, as well as the importance of collaborative problem formulation for data analytics in local governments. This framework could be useful for researchers who are interested in data analytics in local governments to engage in a more comprehensive view and understand data analytics as a socio-technical process and transformational practice, in particular by including the important role of collaborative problem formulation (see Fig. 3).

OPEN IN VIEWER

First, in line with the literature on digital government, innovation, and smartness (Barns, 2018; Gil-Garcia & Sayogo 2016), leadership was found here to be a driving force in setting the direction for data analytics use. As observed, in Syracuse such leadership starts with political commitment with the DA agenda and is supported along the way by the WWC partners (external stakeholders) and City Hall champions (internal stakeholders) that are involved in the different efforts. Since members of the Innovation Team are part of a highly collaborative environment and appear to enjoy autonomy for their responsibilities with the project, leadership in DA could also be interpreted as a transformational force that is triggered by a few actors and stewarded by collaborators both inside and outside City Hall.

Second, as mentioned earlier, problem formulation should be clearly identified as an important initial step for defining what needs to be addressed through DA. It also seems appropriate to link collaboration, a construct that is not new to digital government literature and continues to be studied, to actual problem formulation efforts. Since the research in problem conceptualization and formulation in the context of data analytics is still limited, those theoretical linkages should be further explored empirically. One possibility is to further understand collaborative problem formulation, by explaining which stakeholders and what conditions should be in place to help ensure a more effective DA process in local governments.

As suggested by the Syracuse experience, the existence of ad hoc approaches to DA indicates that some operational flexibility is needed to accommodate multiple ways to define and address any problem with the data. The relatively newness of the topic, particularly in terms of DA practices in local governments, could benefit from more flexibility in defining the models and frameworks for DA use and specifically for actual collaborative problem formulation. Also, more categories or lenses could be added considering new technical or theoretical advancements. New concepts and their relationships should also be considered for revising our proposed framework and, more generally, adding to our knowledge about DA and its impacts on government innovations in both delivery of services and implementation of policies at the local level.