Mapping the organizational design of policy innovation labs: A systematic literature review

We identified three subdimensions describing a lab’s sense of purpose (see Table 5). The first is the interpretation of the purpose of innovation: what is innovation for? We begin by reviewing some existing classifications of the purpose of innovation. McGann et al. (2018) distinguish between incremental innovation, which improves existing practices, and radical innovation, which changes how things are done by reframing public problems, solutions, or processes. Soma et al. (2024) extend this distinction with the concept of transformative incrementalism, arguing that labs are well suited to gradual changes with transformative potential. Lindquist and Buttazzoni (2021), in turn, classify innovation according to goals such as human resource development, the expansion or transformation of government activities, the balancing of operations, and the optimization of results. Other authors draw on the OECD’s (2021) Public Sector Innovation Observatory framework, which identifies four purposes of innovation in public organizations: achieving a specific goal; improving current functions; adapting to the environment; and anticipating challenges or opportunities.

Building on these classifications, our analysis identified three types of innovation purpose: goal accomplishment, performance optimization, and adaptation to the environment. Goal accomplishment frames innovation as a process of causal analysis and the identification of effective tools (McGann et al., 2018; Cole & Hagen, 2024). Performance optimization involves reducing bureaucracy, streamlining administrative work, and improving agency operations (Cole, 2022; Lewis, 2021; McGann et al., 2018). Adaptation to the environment emphasizes integrating external ideas to better align the organization with changing environmental conditions (Komatsu et al., 2021; Toñurist et al., 2017).

The second subdimension is the lab’s theory of change, that is, its interpretation of how innovation is generated. It can be observed in four ways: organizational learning; open innovation; capacity building; and systemic change. Organizational learning draws on March’s (1991) distinction between exploration and exploitation as mechanisms through which organizations generate innovation (Toñurist et al., 2017; Favoreu et al., 2024). Open innovation holds that organizations benefit from involving external actors in decision-making; accordingly, some labs foster innovation through collaborative arrangements such as co-creation and co-production (Brock, 2021; Lee & Ma, 2020; Lewis, 2021). Capacity building assumes that training and related efforts create the conditions needed to strengthen innovation capacity in public agencies (Komatsu et al., 2021). Systemic change, finally, treats innovation as the result of broader institutional, regulatory, or political reforms.

The third subdimension is specialization, which refers to the area in which labs choose to concentrate their efforts. Labs may specialize by geography, type of public agency, policy domain, government area, or methodology. For example, Lindquist and Buttazzoni (2021) classify labs by geographic focus (some target urban planning in a city, while others seek to influence a policy domain at the regional or national level). Labs may also specialize in specific policy areas such as education, health, or public safety (Criado et al., 2021; Fleischer & Carstens, 2022), or in particular methodologies such as behavioral science, digital services, data science, or design thinking (Lindquist & Buttazzoni, 2021). This list is illustrative rather than exhaustive, and future research may identify additional forms of specialization.

Organizational culture gives labs their identity and shapes their activities, communication style, and how they interpret events (Waardenburg et al., 2020; Lindquist & Buttazzoni, 2021; Komatsu et al., 2021; Krogh, 2024). We begin by reviewing some existing classifications of organizational culture. First, design culture aims to bridge the gap between policy design and implementation and promote co-creation (Komatsu et al., 2021). Second, innovation islands refer to labs that operate in isolation from the broader public administration, fostering experimentation without fear of failure (Toñurist et al., 2017). This culture seeks to counteract the risk aversion typical of bureaucracies (Waardenburg et al., 2020; McGann et al., 2018; Torvinen & Jansson, 2023), creating a workspace where participants are less constrained by hierarchies, blame, and risk aversion, encouraging tolerance and patience.

Lindquist and Buttazzoni (2021) identify four types of organizational culture relevant to labs: collaborative clans, creative adhocracies, market competition, and hierarchical control. These cultures are differentiated by leadership style, organizational values, and their approach to effectiveness. Other authors describe additional cultures that support public innovation, such as innovation culture, software development culture, and data culture (Waardenburg et al., 2020; Brock, 2021; Hoss-Golan et al., 2024; Komatsu et al., 2021; McGann et al., 2018).

Building on these classifications, our analysis identified two subdimensions of organizational culture: institutional logics and values (see Table 6). A lab’s institutional logics may emerge from and consolidate through the professional identity and backgrounds of its members (Hoss-Golan et al., 2026; Wellstead et al., 2025). Professionals from fields such as design, software development, and data science play a key role in shaping that identity (Komatsu et al., 2021). It is also shaped by the roles members see themselves performing within the public sector, such as designer, enabler, trainer or mentor, policy analyst, data scientist, software developer, advisor or consultant, and game-changer (McGann et al., 2018; Criado et al., 2021). In turn, professional identity influences organizational culture through members’ mental models, methodologies, practices, jargon, and communication style. For instance, a lab whose members identify as consultants may adopt a more proactive approach to generating projects (Lewis, 2021; McGann et al., 2018).

The second subdimension is values, understood as the shared beliefs of members about the behaviors that enable the organization to achieve its goals. Values are important in labs because they define which behaviors are seen as desirable within the organization, thereby shaping both individual motivations and the selection of lab activities. Based on the public sector value inventory developed by Jorgensen and Bozeman (2007), we identified four main values driving innovation labs: rationality, collaboration, empathy, and creativity. Rationality emphasizes evidence-based solutions and the scientific method as guides for rational public action and planning (Peters, 2020). This aligns with the traditional bureaucratic approach, in which decision-making follows a structured and logical process: examining available information, setting clear objectives, and choosing among alternatives according to quantifiable decision rules.

The value of collaboration refers to involving citizens and stakeholders in government decision-making processes. In some labs, it takes precedence over innovation, with greater emphasis on citizen participation in the design of public services. Empathy is a relatively new public service value that guides user-centered design by focusing on users’ needs and experiences in the creation of products or services. It involves understanding and addressing users’ perspectives to improve their experience. User-centered design methodologies emphasize the human dimension of users, and some authors even suggest replacing the term “user” with “person” to better capture the broader human experience (Villa-Alvarez et al., 2022).

The value of creativity encourages lab members to think beyond conventional approaches in search of novel solutions. It is also associated with the idea of embracing failure, an intervention approach that promotes rapid, low-risk learning and is captured in the phrase “fail fast, fail often.” From this perspective, early failure is treated as a natural part of developing new ideas and as a temporary step toward refinement through iteration. We also grouped several related values under post-bureaucratic values. First, horizontality promotes structures that minimize hierarchy and encourage equality among members, fostering non-traditional forms of leadership (Cole & Hagen, 2024; Asenbaum & Hanusch, 2021). Second, flexibility emphasizes reducing norms, routines, and formal structures so organizations can better adapt to changing environments and collaborators’ needs (Brock, 2021; Villa-Alvarez et al., 2022; Waardenburg et al., 2020).

Finally, some labs are guided by pragmatic values. For example, efficiency emphasizes improving public service performance and, in some labs, becomes the central value, driving incremental change, the optimization of existing functions, and an identity centered on improvement or optimization (Krogh, 2024). Other labs prioritize problem-solving, focusing on rigorous diagnosis and the effective implementation of solutions.

Structure refers to the lab’s organizational form, including its size, functions, chain of command, and other formal attributes. Structure is often established by an institutional mandate, although it can also emerge informally (Bason, 2017). We identified six subdimensions of a lab’s organizational structure: autonomy, authority, position, size, formalization, and funding model (see Table 7).

Autonomy refers to the level of independence a lab has from other public agencies. Labs can be an entirely independent agency or can be subsumed into another agency. Autonomy affects not only a lab’s innovation capacity but also its integration into government functions and long-term survival. For some authors, autonomy is desirable because it allows labs to create a creative, experimental space conducive to incubating innovative solutions (Bason, 2017); greater autonomy often leads to formal or procedural isolation from other public entities, fostering “innovation islands” or safe spaces that enhance innovation capacity (Toñurist et al., 2017; Krogh, 2024). However, this isolation can also lead to detachment from the needs, goals, and culture of other government agencies, potentially reducing a lab’s relevance or effectiveness (Bason, 2017; Cole, 2022).

Authority refers to a lab’s formal ability to influence the operations of other public agencies, providing the most direct means of impact. Without authority, labs must rely on alternative strategies to make themselves relevant in the whole of government, for example, they rely on the persuasive abilities of their leaders to demonstrate the lab’s value to government functions (Toñurist et al., 2017; McGann et al., 2018).

Position refers to the level at which a lab plays a central role or becomes relevant to the whole of government—or rather it plays a peripheral role. We begin by describing how scholars have classified this subdimension. McGann et al. (2018) use two dimensions to classify a lab’s role in the whole of government: formality, indicating whether the lab is formally part of the government; and proximity, indicating its closeness to the executive branch. Then, they classify labs into four types: internal proximate, internal peripheral, external proximate, and external peripheral. Internal proximate labs are closely controlled by the central administration, while internal peripheral labs have greater legal autonomy. External proximate labs are outside the government but maintain formal ties, such as consultants, whereas external peripheral labs engage in informal advocacy or lobbying, like interest groups. Other authors base their classifications on the lab’s legal status (e.g. public, private, public–private partnership, or nonprofit) (Olejniczak et al., 2020; Criado et al., 2021; Lewis, 2021).

Building on these classifications, our analysis identified three types of a lab’s position in the whole of government: centralized, decentralized, and outside government. Centralized labs are formally grounded in the government or maintain close collaborative relationships with public agencies. Decentralized labs, while still aligned with governments, exhibit a lower degree of collaboration or formal association with governments. Labs outside the government may function either as collaborative entities (providing advisory, coaching, or consulting services) or as advocacy groups (created and run by citizens or interest groups), without relationships with governments.

Lab size refers to the number of regular members, including employees and volunteers. Labs typically prefer a lean, horizontal structure with a small core team, prompted by the belief that it enhances a lab’s agility, flexibility, and innovation capacity (Lewis, 2021; Fleischer & Carstens, 2022). Formalization refers to the extent to which rules, routines, roles, and decision-making processes are clearly defined within a lab (Waardenburg et al., 2020). Labs typically have low levels of formalization, operating with minimal rules and structures, as it is expected that they will foster innovation by enabling employees to experiment with new work methods, break away from traditional decision-making processes, and collaborate more easily with external actors (Fleischer & Carstens, 2022).

Finally, labs differ in their funding models, which influence their sustainability, leadership styles, and interactions with other entities (Torvinen & Jansson, 2023). For instance, labs focused on generating revenue through consulting (or project sales) tend to prioritize sales efforts and are primarily accountable to their customers. In contrast, labs that rely on sponsorship are less sales-driven and more inclined to align with their sponsors’ agenda (Whicher, 2021). Following Whicher’s (2021) classification of funding models (sponsorship, contribution, cost recovery, consulting, and a hybrid model), our analysis identified the following categories: sponsorship seeking, consulting, and grant seeking. Sponsorship seeking occurs when the lab receives a budget from a government entity. Consulting involves the lab operating as a consulting firm for other government entities, selling individual projects, making the lab an entity open to multiple types of clients. The grant seeking model refers to the systematic and proactive lab’s effort to receive funding through grants from national or international organizations.

In this dimension, we list and classify the tools innovation labs use to intervene and drive change, beginning by highlighting existing classifications. Lewis (2021) categorizes them into three main approaches: user-centered design, evidence-based decision-making, and process streamlining. Criado et al. (2021) classify the tools into design, data, behavioral economics, and hybrid categories, while Lee & Ma (2020) identify similar groupings, including experimentation, data science, evidence gathering, and evaluation. Other lists echo these categories, mentioning design thinking, ethnography, field research, experimentation, artificial intelligence, hackathons, and big data (Wellstead et al., 2024), as well as behavioral economics, co-production, system dynamics, and randomized controlled trials (Whicher, 2021). It is important to note that a single instrument or group could fulfill other purposes (e.g. enhancing public service experience, fostering citizen participation, or streamlining administrative procedures).

Building on these classifications, our analysis identified five groups of instruments: user-centered design, experimentation, iterative development, collaboration, and data analysis (see Table 8). First, user-centered design is an approach that emphasizes involving the end-user as an active participant in the artifact design process (Bason, 2017; Torvinen & Jansson, 2023; Villa-Alvarez et al., 2022). While design thinking is the primary methodology used to implement this approach, other tools also apply the principle, such as ethnography, customer journey mapping, experience design, and service design.

Second, experimentation is an approach based on the principle of evidence-based decision-making and rational planning. To produce reliable evidence, labs have adopted innovative tools that lend a scientific quality to the information generated (Soma et al., 2024; Favoreu et al., 2024). Third, iterative development or prototyping is a cyclical design process centered on the design–develop–test triad. It begins with minimal investment in development time and infrastructure (e.g. simple sketches) and prioritizes early user interaction. The process involves refining designs based on frequent feedback from future users. Iterative development is believed to reduce costs, mitigate risks and risk aversion, and accelerate public service design (Torvinen & Jansson, 2023; Cole & Hagen, 2024; Olejniczak et al., 2020). Like experimentation, this approach emphasizes the use of pilot projects, so that performance can be assessed through the scientific method and earlier in the policy cycle. Unlike experimentation, which evaluates predefined problems and policies, iterative development is a design tool used to explore and define public problems and policies, and to detect implementation failures before scaling up (Peters, 2020).

Fourth, instruments for collaboration are related to the use of the frameworks of co-creation, co-production, crowdsourcing, and participatory budgeting, and tend to overlap with complementary government agendas, such as fostering citizen participation or strengthening collaborative governance (Asenbaum & Hanusch, 2021; Torvinen & Jansson, 2023; Favoreu et al., 2024). Finally, data analysis is an approach leveraging technological innovations, particularly the emergence of big data, data science, and artificial intelligence (Kim et al., 2023; Hoss-Golan et al., 2024). These instruments help strengthen public policy analysis through better data and more sophisticated quantitative analysis techniques.

Labs have different strategies for attaining their defined mission and scope. We begin by highlighting existing characterizations of the lab’s strategies. Several authors have observed that labs organize their strategy around different activities, goals, or functions: analyzing public problems, creating solutions, exploring situations, testing solutions, and redefining policy issues (Cole & Hagen, 2024); conducting user research, co-designing with the public, co-prototyping, public consultation, and piloting or scaling solutions (Olejniczak et al., 2020); reframing problems, designing and testing prototypes, reforming public services, and developing public policy (McGann et al., 2018); problem recognition, idea generation, prototyping, and scaling (Criado et al., 2021); consulting, training, intervention, research, and prototyping (Lee & Ma, 2020); user research to redefine policy, stakeholder engagement for co-design, project prototyping, consulting, and monitoring or evaluating public policy (Whicher, 2021).

As many of these categories overlap or lack a clear definition or reference to specific actions, we sought to narrow the listing of functions to those with clear definitions and differentiation. Taking this into consideration, we organized the subdimensions of strategy in the lab’s function, participant selection, and policy stage. Beginning with the lab’s function, our analysis identified three types of a lab’s function: research, creation, and knowledge transfer (see Table 9). We identified three types of function: research, which includes generating information, reports, and conducting user surveys or idea generation; creation, which covers the design, ideation, and prototyping of solutions; and knowledge transfer, which encompasses training, mentoring, advising, coaching, or consulting activities aimed at public officials.

The second subdimension is the participant selection process, which organizes the form of interaction and contribution among participants in the innovation processes. Participant selection occurs through deliberate strategies, or owing to constraints and limitations. For instance, Krogh (2024) notes that public officials or a broad spectrum of potential users are sometimes excluded. McGann et al. (2021) emphasize that middle and senior managers are often left out, which impedes implementation. Including them can ensure that these managers understand the rationale and evidence behind collaborative proposals and help refine services based on their expertise. Adding to this complexity is the strained relationship between experts and non-experts in deliberations, as well as the tension between innovation and collaboration.

Finally, the third subdimension is the policy stage in which the lab operates. Labs may focus their activities on one or more stages of this process. For instance, they may influence the public agenda by generating research reports and publishing data analytics (Kim et al., 2023); contribute to problem definition through interviews and ethnographic research with users (McGann et al., 2018); assist in policy formulation using design thinking, co-creation, and other methods to develop solutions (Criado et al., 2021; Favoreu et al., 2024); or support implementation through co-production and policy scaling. Some labs also specialize in data analytics to enhance public policy monitoring (Kim et al., 2023).

The literature suggests that labs primarily focus on the public policy definition and formulation stages (Whicher, 2021; Olejniczak et al., 2020), as they enable small-scale experiments that are both cost-effective and agile through iterative development or prototyping. However, Villa-Alvarez et al. (2022) argue that many labs also impact the implementation stage.