Barriers to innovating with open government data: Exploring experiences across service phases and user types

2.1 Non-pecuniary outbound open innovation from OGD

OGD is the subset of open data that is produced by publicly owned organizations. Two distinctive features make the release of OGD particularly promising. Firstly, public organizations create and collect vast amounts of data in many different domains (Janssen et al., 2012). Thus, they hold potential to single handedly create a large solution space for innovation. Secondly, as an integral part of public operations, taxpayers have paid for the creation and collection of OGD. Accordingly, compared to other forms of open data, OGD constitute a shared resource that offers value beyond what is captured from the original intended use (Jetzek et al., 2014).

While acknowledging that OGD also is used directly (cf. Davies, 2010), we argue that services or products that transform the data into practical knowledge is a cornerstone in making OGD useful for citizens. Thus, in our view, aggregation and refinement of OGD is required to reap the proposed societal value. We consider this activity as innovation, that is “an intentional and proactive process that involves the generation and practical adoption and spread of new and creative ideas, which aim to produce a qualitative change in a specific context” (Sørensen & Torfing, 2011, p. 8). Further, we also argue that fruitful cultivation of OGD implies that public organizations must look beyond their organizational borders and involve external actors in the innovation process. Hence, we adopt an open innovation (OI) lens to position the case study.

Chesbrough (2003) identified a paradigm shift away from closed innovation models, and denoted the movement OI. He moreover summarized its emphasis as “the use of purposive inflows and outflows of knowledge to accelerate internal innovation, and expand the markets for external use of innovation, respectively” (Chesbrough, 2006, p. 2). Gassman and Enkel (2004) explored three types of openness: inbound, outbound and coupled (see Fig. 1). In the inbound form, the focal organization opens its organizational borders to purposively insource external ideas and technologies in order to enhance its products’ values. In the outbound form, the initiator instead outsources its internal resources, and allows external actors to refine, exploit and bring them to market. The third form is labeled coupled. It is a combination of the inbound and outbound processes, often characterized by the formation of enduring alliances with complementary external partners. Dahlander and Gahn (2010) moreover distinguished between pecuniary and non-pecuniary forms of OI, that is whether or not the initiator seeks to reap immediate financial rewards or indirect benefits.

OPEN IN VIEWER

The release of OGD to accelerate innovation from distributed OGD users is grounded in the strategy of revealing internal resources to the external environment. These internal resources are generally made accessible without any immediate financial compensations, but with the hope of indirect benefits. Therefore, we consider such practices as non-pecuniary outbound types of OI. Accordingly, we position our case study as a non-pecuniary outbound OI in the public sector, focusing on the barriers that OGD users experience when innovating with OGD.

Although outbound OI from OGD is promising and theoretically sound, the practical impact so far remains limited. Both OGD providers and users face a number of barriers in generating value. Extant research on open data has often either focused on barriers for providers, users or the role of intermediaries in mitigating them. For example, Barry and Bannister (2014) propose a typology of barriers for data providers, Zuiderwijk et al. (2012) identify 118 impediments and emphasize how open data policies neglects users despite them being the actual value creators, Kaasenbrood et al. (2015) and Maccani et al. (2015) both identify factors influencing the adoption of open data by private organizations, Hjalmarsson et al. (2014) provide a framework for barriers experienced in relation to OI contests, and, Smith et al. (2016a) as well as Jetzek (2016) both suggest design considerations for open data intermediaries. In a comprehensive working paper, Ubaldi (2013) moreover shows that successful open data initiatives require a holistic view. The paper lists policy, technical, economic and financial, organizational and cultural, as well as legal challenges.

For the purpose of this study, the classification of innovation barriers suggested by Jansen et al. (2012) is adopted as it provides an appropriate level of abstraction and has proved useful in a large number of studies. In this classification, six interrelated categories of barriers towards open data provision and adoption are identified: institutional, task complexity, use and perception, legislation, information quality, and technical.

In sum, extant research has identified, described and categorized a large number of innovation barriers. Still, the issue of how their impacts vary, between different types of OGD users and across their services’ lifecycle phases, remain largely uncharted.

2.3 OGD users

Extensive research on why people contribute to open source projects has in short come to the following conclusions: motivations differ extensively between individuals (Juell-Skielse et al., 2014), the ‘open’ ideology fuels development (Shah, 2006), open source developers are both intrinsically (e.g. Ghosh, 1998) and extrinsically motivated (e.g. Kuan, 2001), participation is often driven by factors similar to other social movements (Hertel et al., 2003), and the motivations varies over time (Shah, 2006). In contrast, research on motivation among open data users is scarce. In a rare study, Davies (2010) finds that they are mostly male originating from smaller businesses in the private sector, the public sector or academia. Davies (2010) moreover identifies six sets of drivers for open data use: understanding the government, creating technological innovations, seeking recognition or profit, improving government through digitalization, solving specific problems and providing novel public sector services. Antikainen et al. (2010) discover that contributors to OI communities value intangible factors such as community cooperation, support, skills development and amusement, over monetary gains. Ståhlbröst and Bergvall-Kåreborn (2011) add that open data users’ motivation is influenced by both community type and technology adoption type. Zuiderwjk et al. (2015) state that performance expectancy, effort expectancy, social influence, facilitating conditions and voluntariness of use together account for 45% of variability in intention to use open data (for any purpose). Nevertheless, thorough descriptions of open data users are missing in general, and in particular little is known of who the OGD users are, what motivate them to develop services and how they go about for doing so. Moreover, the understanding of how their characteristics correlate with their experiences of innovation barriers is low. Consequently, in this case study, we first set out to identify a topology for different types of OGD users, in order to enable an analysis of how their characteristics affect perceptions of innovation barriers throughout their innovation processes.

4.1 Experiences of barriers throughout the service lifecycle

In the following section, the perceived innovation barriers in each phase of the service life cycle are outlined. A summary of the main barriers experienced in each phase is provided in Table 1, while the full list of 38 identified barriers can be found in the appendix.

Strategy phase. OGD users experienced difficulties in communicating with, understanding, influencing and getting help from OGD providers and potential business partners. Thus, it was difficult for them to foster and nurture service strategies where they could control their project processes and outcomes. The absence of service level agreements (SLAs) moreover made it difficult to establish sustainable business models. This was partly due to uncertainty on the long-term consistency of the data provision, but also because they could not guarantee a level of service to the end-users of their services. The lack of SLAs impeded end-users’ willingness to pay, an issue that was particularly detrimental as the potential end-user pool was perceived as limited, the digital service market as fairly saturated and the general cost acceptance among end-users as low.

The quality is vital, and they still lack a SLA for that regarding this data, or for any data that goes through Trafiklab that I know of. There is no guarantee. Instead it’s best effort. Best effort is damn hard for an individual or a small business to make money from. It’s only hobby projects that will be pursued, if they cannot guarantee the quality. – IP16 (translated)

Design phase. The most prominent barrier during the service design phase was the overall lack of OGD. OGD users requested more, richer and increasingly diverse data in order to broaden the range of possible services. 17 out of 19 interviewees emphasized this issue. The cumbersome registration process and the difficulty in getting an overview of the APIs’ content further augmented the perceived lack of data access.

Trafiklab provides public transport data. At some point, you run out of ideas. You cannot do much more than present the public transport. I believe that, in order to get a little more dynamic in the data they provide and somewhat greater diversity among the apps, or whatever innovation one can develop, they need to take in more data, but still data that concerns traffic. – IP18 (translated)

Task complexity barriers were likewise noticeable during the design phase. Designing digital services require a wide spectrum of skills, such as developing system architectures and service wireframes (i.e. skeletal frameworks of the user interfaces). As a result, many of the often programming-oriented OGD users were forced out of their comfort zones. OGD users less acquainted with digital development were on the other hand often repelled by the technical and opaque appearance of the marketplace.

The format of the APIs provided by the studied OGD marketplace was a major barrier, hindering OGD users from designing services as they wished. It did not permit downloading raw bulk data or subscribing to information updates. The effect of this was twofold. First, the format did not facilitate system architectures that stock parts of the intelligence internally – increasing OGD users’ dependency on data provision reliability and lowering the scalability of the ecosystem (see service strategy barriers). Second, the lack of access to raw data moreover restricted the OGD users to design services within the OGD providers’ envisioned scope.

If I provide a product like a spoke wrench for bikes, you cannot do much else [than adjusting and replacing spokes]. But if I provide a hammer, you can suddenly use it for much more than I originally intended. …You might create artwork, take down a wall, or whatever. – IP5 (translated)

Transition phase. The transition phase was home turf for a majority of the OGD users. Yet, the lack of documentation, example code, support, and test facilities added intricacy to the task. Several expressed frustration with having to deal with issues related to data access at the expense of the problems they actually are interested in, for instance making use of the data. The task complexity was further augmented upon nonexistent standards, fragmentation of resources and lack of mapping tables. This made it hard to combine information from several APIs. As both the community’s functionality and activity was fairly low, the OGD users also had a hard time trying to benefit from others’ knowledge or experiences to overcome these impediments.

Many station names are cut off, there are columns with false data and mapping tables are missing between different APIs. Therefore, you have to develop such mapping tables yourself if you want to combine different types of data. – IP4 (translated)

Another salient barrier category was use and participation. OGD users struggled to motivate, find and finance the required time for service development. As writing the code represented the lion share of many of the projects, the time barrier was particularly apparent during the service transition phase. However, some also struggled with the time for publishing the services, especially as many predicted the monetary benefits to be small.

At the minute, time is a major barrier for me. This is something I at least must develop into a proof of concept on my own time. Then I might be able to show it to my company, and they can then feel that; this is a good idea, it’s within our sphere and we can invest in it. – IP11 (translated)

Most OGD users moreover experienced problems with establishing and communicating unique value propositions for both end-users and business partners. Thereupon, they were stuck in a chicken or egg type of dilemma – not able to establish either the partnerships needed in order to reach end-users or end-user bases that would make the services attractive for potential partners. Overall, the OGD users found it very difficult to market their services and to reach end-users.

Sure, there’s a huge need [for my service]. The only problem is to reach out in the roar of 1 1 / 2 million other apps available on App store. – IP1 (translated)

Operation phase. The operative phase was hampered by the slow data provision and the low back-end reliability. These barriers made it hard for OGD users to deliver accurate information in time through their services. This issue was particularly troublesome since the end-users tended to blame the OGD users when the service quality was low, regardless of the quality of the data provision. The lack of communication and poor support from OGD providers moreover obstructed the OGD users’ abilities to understand current and future statuses of the data provision as well as to solve all the various issues that came up during the operation.

If you have a product, which is up and running, but then stops working – then you don’t want to send a letter with the post. It takes a lot of time and you don’t know if the receiver checks his or her post box. He might even be on vacation, and then you don’t know who to send the letter to. They ought to be more active in some way, as they gain extremely much from things that we develop for free. – IP5 (translated)

Continual improvement phase. End-users were perceived to supply very little feedback in general, and the critique that came through to OGD users was seldom constructive. Thus, it was difficult for the OGD users to assess the services’ level of usability and performance, that is to understand how the services could be improved. It was moreover difficult for the OGD users to motivate putting effort into updating their services, since publishing them often was the ultimate goal of their projects. Furthermore, very few of the OGD users earned any money from their services. So, for most OGD users there were no monetary incentives either.

The barrier for me is that I only do the things for myself, because I enjoy tapping on the keyboard and see if something happens. I’m not an entrepreneur. I don’t find it enjoyable to turn it into a product and launch it. I usually only do things half-done, as it’s only in the beginning it’s fun. – IP9 (translated)

The 19 interviewed OGD users were categorized as employee, entrepreneur or hobbyist. The categories differ in motivations, objectives, pre-conditions and approaches as well as in experiences of innovation barriers, see Table 2.

Employees. OGD users categorized as employees utilized OGD as a part of their employment, in order to fulfill pre-defined work tasks. In relation to the OGD marketplace, their tasks were often to integrate information from the provided APIs into existing or developing services. The employees commonly worked as developers for IT consultancy firms that deploy several projects simultaneously, but also worked in various roles for businesses that focus on single projects. The employees were primarily driven by the outcome of their projects, that is to fulfill managements’ and/or clients’ expectations. Thus, they were generally fairly focused and goal-driven – predominantly concentrating on getting access to and decoding the data they need. Therefore, they neither valued nor participated in knowledge sharing activities. Seven of the 19 interviewees were catalogued as employees.

On average, the employees dwelled less on barriers, compared to the two other OGD user archetypes – both in terms of quantity and emphasis. However, being solely focused on getting hold of the OGD, the employees’ major concerns were the scope and quality of the data provision service. They tended to expect the same professional approach from OGD providers and intermediaries as they experience from their other suppliers, partners and clients. Hence, they were surprised and disappointed with the level of help in terms of documentation, example code and support. It impeded their efforts of designing and developing services. Lack of data and the data format moreover hindered them from both developing strategies and designing services as they wished.

Overall, employees’ innovation efforts and their services’ lifecycles were a lot less concentrated around and dependent on the innovation ecosystem surrounding the studied OGD marketplace. Instead, their ideas, project financing, business partners and feedback frequently originated from other innovation ecosystems.

Entrepreneurs. OGD users categorized as entrepreneurs utilized OGD to pursue service development projects aimed at either monetizing or at gaining attention and reputation. Their projects were more often than not based on a combination of an identified gap in the existing range of services and a novel service opportunity offered by the data provision of the OGD marketplace. While their primary focus commonly was on one project at a time, the entrepreneurs often had several project ideas in their pipelines, awaiting the right timing to pursue them. Both internal and external driving forces motivated them, that is the challenge of developing services and businesses as well as the potential gains of the project outcome. The entrepreneurs were generally interested in all the OGD marketplace’s activities, where they often used knowledge-sharing activities to promote their projects and themselves. Five of the 19 interviewees were catalogued as entrepreneurs.

The entrepreneurs were the most outspoken OGD user category, having endless of experiences, opinions and suggestions regarding the OGD marketplace that they wanted to share. As the employees, entrepreneurs were limited by the lack of data and the lack of support. However, the entrepreneurs were to a greater extent also hindered by the inherent complexity of the task, as they often were operating on their own and not always as acquainted with writing code as the other user types. They frequently mentioned struggling with making sense of the fragmented documentations and overviewing what information the APIs contained. Moreover, the entrepreneurs also had bigger problems with allocating time during all phases of the service lifecycle, compared to the other archetypes.

Institutional and legislation barriers were particularly detrimental for the interviewed entrepreneurs. Many of them developed services that use the OGD marketplace as the core data provider. As a consequence, their services were highly dependent on the data provision of the OGD marketplace, but as a minor player, they still experienced immense difficulties when trying to gain insight to and influence the activities of the OGD providers.

Hobbyists. OGD users categorized as hobbyists utilized OGD to pursue their hobby projects. Internal driving forces typically motivated them. Their primary objectives when interacting with the OGD marketplace were thus often to explore the published data, to have fun and to solve problems. These problems mostly related to either the hobbyists’ everyday issues or to purely technical challenges. The hobbyists pursued several smaller development projects simultaneously, often at an irregular pace. Most of these projects will never be finished and the hobbyists’ objectives were seldom to publish the services they developed, as they predominantly were interested in the early phases of the service lifecycle. Seven of the 19 interviewees were catalogued as hobbyists.

The hobbyists were very different from the other two types of OGD users in terms of objective and approach. Being mainly interested in exploring the data, ‘playing with it’ and using it to solve self-experienced problems, the hobbyists were not too concerned about matters such as contractual issues or about the barriers that appear during the later stages of the service lifecycle. Further, being both more flexible and more programming-oriented than the other OGD user types, the hobbyists were not as frustrated with specific technical barriers during the transition phase. However, the hobbyists were also troubled by the lack of data and struggled with finding time, just as the entrepreneurs.

All interviewed hobbyist developed services that used the OGD marketplace as the core data provider for their services’ functionalities. Still, being less focused on publishing their services, institutional and legislation barriers were not as central for them, as for the entrepreneurs. Instead, the hobbyists spoke more about the underpinning values of disclosing OGD than the two other groups. As a consequence, their frustration in regards to tedious authorization procedures and limited data sets was higher. Lastly, the hobbyists were in general also more interested in co-creation and discussions with other OGD users. Thus, the lack of community functions was particularly unsatisfactory for them.

5.1 The dynamic effects of innovation barriers

Extant research has identified a vast number of innovation barriers that impede release and adoption of OGD. However, how effects of innovation barriers vary across distinct OGD user types and activities in innovation processes has received little attention. To this end, we studied a specific set of OGD user, service developers. We conducted an inductive analysis to identify distinct types of OGD users in this group, and we traced how effects of innovation barriers vary across OGD user types and across phases in their innovation processes. Our findings provide a number of insights on the dynamic effects of innovation barriers.

The analysis reveals that barriers hinder OGD users’ innovation efforts throughout the entire service lifecycle. This finding challenges the use of open data production, distribution and consumption (e.g. Zuiderwijk & Janssen, 2014) as model when considering OGD impediments. Instead, it suggests that also later service lifecycle phases must be considered in order to facilitate throughput in terms of innovation. Moreover, the analysis illustrates that OGD users face different types of barriers in different lifecycle phases. For instance, the lack of data, the impeding data format and the high task quality were perceived to be most challenging in the design phase, while the operation phase rather was impeded by lack of communication from data providers, poor support, slow data provision and low back-end reliability. This finding suggests that an analysis of which phases that primarily hamper OGD users in a given innovation ecosystem can inform the process of choosing appropriate measures for catalyzing its throughput. Several of the identified innovation barriers, such as lack of transparency and influence are also social rather than technical in nature. Hence, the analysis enforces earlier reports that the mere provision of data does not generate value on its own (e.g. Janssen et al., 2012), and highlights the importance of going beyond technical aspects when ensuring that the possibilities for translating OGD into useful outcomes is available for the widest possible range of OGD users (cf. Gurstein, 2011). Furthermore, the interviewees expressed experiences of a multitude of innovation barriers that are not directly related to the OGD provision. Examples include lack of domain knowledge during the strategy phase, impeding client requirements during the design phase, insufficient programming skills during the transition phase, time constraints during the operation phase and lack of feedback during the continual improvement phase. This finding emphasizes the importance of utilizing a holistic innovation ecosystem perspective (cf. Adner & Kapoor, 2010) when analyzing OGD users’ possibilities to reap benefits from OGD.

The analysis of OGD user archetypes extends earlier findings on the diversity among the service developing crowd (e.g. Zuiderwijk, 2015) and variations in reasons for using OGD (e.g. Davies, 2010) as it provides a more granular examination of these issues. The paper portrays how OGD users differ in terms of motivations, pre-conditions, approach and objectives, and suggest three archetypes based on these criteria: employees, entrepreneurs and hobbyists. These character traits suggest that the different types of OGD users can play different roles in the evolvement of OGD innovation ecosystems (cf. Styrin et al., 2017). For example, due to their motivations and goals, entrepreneurs seem more likely to pursue high profile projects utilizing provided OGD as core. These projects can be used as best practice and for increasing the public awareness of the innovation ecosystem as well as the perceived attractiveness of participating in it. Employees, on the other hand, have unique abilities to integrate the provided OGD into solid services that origin from other innovation ecosystems. Thereupon, they are also important for the expansion of the ecosystem. Lastly, being data-driven, focused on multiple projects, and interested in knowledge sharing activities, hobbyists are important for acquiring continuous activity and cross-border knowledge transfer in the innovation ecosystem (cf. Smith et al., 2016b).

This study also adds to previous literature on impediments for OGD use (e.g. Zuiderwijk et al., 2012) by demonstrating that the OGD users’ motivations, pre-conditions, approach and objectives have an impact on the experiences of innovation barriers. Motivation: Employees, driven by extrinsically factors, were in general less focused on the process as such. Thus, impediments affecting the experience of it, for instance the unappealing appearance of the data provision website, were less hampering for them. On the other hand, compared to the intrinsically motivated hobbyists, both the employees and the entrepreneurs regularly put higher emphasis on impediments affecting the solution space for services, such as the limiting data format. Pre-conditions: The employee driven projects were pursued as part of principal occupations. The employees’ commitment was therefore, at least partially, ensured through their employment contracts. As an effect, the employees were less likely to be hampered by internal use and participation barriers than hobbyist and entrepreneurs, who often developed services on their spare time. Approach: The employees had clearer guidelines and frameworks for their projects, compared to entrepreneurs and hobbyists who often pursued project upon opportunity. Hence, their flexibility in overcoming technical barriers such as nonexistent standards was in general lower. Objective: Both the business oriented employees and the entrepreneurs were more likely to be hampered by legislative barriers, such as lack of SLAs, than hobbyists. This was probably largely due to their strong outcome orientation and expectations on extrinsic rewards, while hobbyist were less committed to publishing and reaping rewards from their services.

5.2 Implications

For policy-makers and practitioners, the analysis of experiences among the three OGD user archetypes implies that OGD stakeholders such as OGD providers and policy-makers can influence the mix of OGD users in the innovation ecosystem by choosing which types of barriers to work on. For instance, efforts aimed at lowering task complexity barriers during the transition phase can substantially enhance the attractiveness of participating in the ecosystem for employees and entrepreneurs, but would probably not affect the appeal for hobbyists, as they mainly struggle with information quality barriers during this phase. Likewise, use and participation measures would probably impact the willingness and possibility to participate for entrepreneurs, but would not affect employees, as their devotion is regulated by their contracts.

As discussed earlier, different types of OGD users can play different roles in the evolvement of the innovation ecosystem. Innovative developments moreover often happen at the intersection of fields (Hargadon & Sutton, 1997) when innovators manage to connect seemingly unrelated fields, problems, or ideas. Diversity as a driver of innovation has been referred to as the ‘Medici effect’ and points to how innovation often arises from combinations of knowledge and experiences across contexts rather than from isolated ingenuity (Johansson, 2006). A diverse crowd with large coverage provides the ability to turn distant search into local (i.e. leveraging specialists from a wide set of contexts without associated costs being incurred), and to assure the sustainability and innovation potential of the ecosystem. Thus, the palette of actions aimed at attracting OGD users to OGD innovation ecosystems should be organized to support diversity.

The diversity argument is also relevant for the data provision as such. For example, the utilized data format and the setup of the studied marketplace were perceived to hamper what type of services that could be developed. In order to attract a large and diverse crowd and to ensure scalability, it is important to expand the solution space and to lower the dependability as much as possible. For this reason, flexible, call-based data formats and a system design that allows for bulk-download of data seems preferable.

Furthermore, the cross-tabulation of innovation barriers and the service lifecycle phases suggests that measures aimed at boosting non-pecuniary outbound OI practices in the public sector based on OGD should not exclusively focus on initial and technical data provision impediments, although these barriers are highly important during the design and transition phases. Other types of innovation barriers impede OGD users’ ability to sustainably develop, launch and maintain services, thus calling for supplementary actions.

For open government and open data researchers, this paper calls for increased focus on the innovation ecosystem for OGD users. Previous literature on OGD enabled innovation time and again fail to acknowledge OGD users’ reality in the sense that it merely focus on the data supply side of their innovation processes, ignoring their relations with end-users, clients and business partners. In other-words, less data centric and more user centric OGD research is needed.

5.3 Study limitations

This study was confined both geographically (Sweden) and to a particular domain (public transport). Additional studies are needed in order to establish the generalizability of the findings and for widening the understanding of OGD users’ experiences of barriers when innovating with OGD. Moreover, further comparison with studies on OI in the public sector in general (e.g. Feller et al., 2011; Gascó, 2017; Mergel, 2015; Mergel & Desouza, 2013), and barriers to such practices in particular (e.g. Mergel, 2017; Smith et al. 2018), is needed in order to establish if the findings are transferable to other forms of non-pecuniary OI in the public sector.

The outcome of the study is naturally also influenced by the chosen research method. The choice of adopting a categorization for barriers from Janssen et al. (2012) ruled and limited the scope of the analysis. The appropriateness of the ITIL service lifecycle model (Taylor, 2007) as a framework for describing the innovation processes of OGD users also needs further assessment. Despite associating the expressed barriers with different phases of the service lifecycle, it should moreover also be emphasized that some of them are present during several lifecycle phases.

Lastly, the boundaries between the proposed OGD user categories are blurry. Moreover, they neither represent the full spectrum of OGD users, nor display the full complexity in terms of relevant aspects. As such, the categorization merely provides a sensitizing device (Weick, 1976) that illustrate that differences within the OGD user group exist, and that these differences, among other things, impact their experiences of innovation barriers.