The last mile of policy design: the case of urban food waste intervention

Abstract

This study investigates the dynamics and challenges in the specification of policy design at the micro level, when broad policy issues are translated into specific interventions. By conducting an in-depth case study of urban initiative dealing with food waste in Warsaw, we unravel the complex process of translating policy objectives into action. Our findings have three points. First, the dynamics of micro-level design provide a broad space for policy designers to make decisions that reshape both the understanding of the policy problem and the change strategies used to address it. Second, policy design can be both engineering and bricolage. It results from changes in context, political considerations, research evidence, and institutional actors’ pragmatic decisions. Third, the Theory of Change can be used as canvas to guide designers through this dynamic process. Future research is recommended to generalize our findings across different contexts and explore the roles of diverse stakeholders in the policy design process.

Keywords:

• Theory of change
• policy design
• food waste
• urban intervention
• policy lab

1. Introduction

There has been growing interest in public policy design in recent years due to its potential for creating more effective, efficient, and equitable solutions to societal problems (Howlett and Mukherjee 2020). During policy design, policymakers develop assumptions about the nature of the policy problem, what instruments to choose, how, and why these instruments are expected to lead to desired outcomes. This set of assumptions, often operational, has a substantial impact on the actual success of policy in addressing specific societal problems.

The literature discusses a broad spectrum of decisions that policymakers face when designing interventions. Public policy researchers have pointed out the importance of a better understanding of the policy context and matching design activity to its dynamics (Howlett 2018). Colebatch (2018) argues that policy design is an exercise in choosing meaning and framing to make appropriate and valid practices and outcomes. The evaluation literature adds to this importance of understanding causal mechanisms and collecting evidence on what interventions work, for whom, and in what context. It can help organizations make strategic decisions about program’s elements, providing them with a structure for learning and accountability (Pawson and Tilley 1997; Rogers and Williams 2006). Finally, the recently emerging stream of literature on behavioral science provides insights into the mechanisms of both individual and collective behavior and the choice of policy tools aligned with human cognitive and behavioral patterns (Chater and Loewenstein 2022; Weaver 2015).

Despite the rich literature, there are knowledge gaps in the current understanding of policy design. First, there is an emerging critique that most current literature focuses on ideal types and stages and is divorced from a dynamic, iterative reality (Cairney 2021). It seems that real-life policy design is more bricolage than engineering. It includes bargaining and compromise and is not limited to technocratic exercises (Peters and Fontaine 2022).

Second, there is a set of under-examined issues of how groups of policy targets are chosen, how policy designers discover and decide upon policy mixes (Capano and Howlett 2020) and how stakeholders arrive at a shared understanding of new intervention.

Third, policy research has only recently begun to address the micro-level of policy design. The current special issue is one of the first attempts to conceptualize micro-level policy design in terms of specification and calibration (see: Introduction to this volume by Capano & Howlett).

Thus, based on these recent literature developments and building on the issues initially framed in the introduction to this journal, we put forward following question: What does the specification of micro-level policy design look like regarding its merit and the process dynamics?

The article addresses this question with an in-depth analysis of the case study of an urban policy initiative dealing with the food waste problem. We focus on food waste because it exemplifies a complex policy issue that confronts policy designers with multiple choices. It links human-technology-environment interactions. It is also multifaceted; it can be addressed from the perspective of different target groups, sectors, and levels of public institutions (Reynolds et al. 2020).

Our analysis focuses on the specification of the policy design, leaving out the calibration issue. This is because the case study was an experimental intervention developed in the policy lab and focused on the specification and field test of the intervention. Calibration issues related to scaling up and mainstreaming interventions were not covered by this process.

Our research question is exploratory in nature, which corresponds to the initial stage of development in studies of the micro-level of policy design. We hope that our article contributes to a better understanding of how micro-level policy design works in real life. For academics, such mapping could establish a descriptive base for future comparative and explanatory studies, whereas for practitioners, our findings on the design process could help make more conscious choices among quickly emerging options.

2. Research design and methods

The article is based on a single case study, a design suitable for in-depth exploration of the specific phenomenon and the dynamics of processes (Yin 2018). It describes participatory action-oriented research (Bryman 2012). Participation refers to our engagement in co-designing the policy solution in partnership with practitioners from the City of Warsaw, community members, and stakeholders. Action orientation means that the research’s primary purpose was to develop a policy solution that addresses the needs of the local community. The term "research" refers to the soundness of methods used in the process as well as structured and critical meta-reflection.

In practice, this approach means that we, the authors of this article, were operating on the two levels. The first level means our direct involvement in the whole design specification as design team members. We participated in most design activities. The methods used for the design specification included workshops with stakeholders, brainstorming sessions, data analysis of available statistics on policy problems and policy users, surveys with potential users, field observations, and interviews with stakeholders.

The second level means our critical meta-reflection on the co-design process we experienced. We used the reflective practitioner model (Schön 1983) as guidance. For data collection and analysis, we used weekly field notes, structured discussions of personal experiences, and retrospective analysis of project documents.

The case study of the Warsaw Food Lab was built on three conceptual pillars. As an overarching frame for mapping the content of design specifications, we applied the work of Capano and Howlett, which conceptualizes specification in terms of three elements (see Introduction to this volume):

1. Target population: Who is affected by the intervention?

2. Expected outcome: What should be done with respect to the actual situation?, and

3. Time frame for intervention: In what time frame should the intervention be done?

To trace the activities of the designers during the design specification, we used the REACT framework (see Figure 1). The framework acknowledges that designers’ actions are spread between action versus research and theorizing versus diving into complex reality. The main types of activities are: Re-define (make choices on specification elements), Explore (collect data, observations, insights), Analyze (analyze collected data and digest insights), Create (develop interpretations, understanding, solutions), and Test (verify ideas, decisions, and solutions); thus, R-E-A-C-T. The framework has been developed as a result of mapping design processes in policy labs worldwide (Olejniczak, Borkowska-Waszak, et al. 2020), and has already been used by other Polish municipalities during design processes. Thus, it fitted the basic organizational arrangements of our case study – Warsaw food lab.

Figure 1. Map for the process of specification.

Based on Olejniczak, Borkowska-Waszak, et al. 2020.

Finally, we used the Theory of Change to articulate the content of the actual policy intervention, a product of specification. In essence, the Theory of Change (ToC) outlines causal linkages within a policy intervention to explain how and why a desired change is expected to occur in a particular context (Brest 2010). In our case study, the leader of the European consortium clearly indicated that the Theory of Change should be a product of the design process. The leader provided flexibility in choosing the form of ToC by indicating only some key elements such as problem and context, intervention, output, early/longer-stage goals, and assumptions that connect the transition from shorter to longer term goals (source: FT Consortium Workshop on ToC, April 2022). We adapted the ToC to focus on the target group and its behaviors (for a detailed description, see Appendix 1). Theories of Change are a standard tool in modern public policy. They are extensively used by policy practitioners (Simeone et al. 2023; Lemire, Porowski, and Mumma 2023; Koleros, Adrien, and Tyrrell 2024).

3. Case study of food waste intervention in Warsaw

3.1. Background of the Warsaw Food Lab

The design process discussed in this paper was conducted as a part of the project implemented by the City of Warsaw, called The Food Trails Project. The project was co-founded by the European Union (Horizon 2020) and implemented in 11 European cities between 2021 and 2024 (https://foodtrails.milanurbanfoodpolicypact.org/). It aimed to create a paradigm shift in urban food policy by empowering cities to reconfigure and implement sustainable, healthy, and inclusive food policies, instigating change from the grassroots to the institutional level. A crucial component of this initiative were "policy labs" - spaces for dialogue, collaboration, and co-design with stakeholders of urban food systems. Cities were expected to create and manage such labs. This idea is related to similar policy innovations worldwide (Hossain, Leminen, and Westerlund 2019; Wellstead, Gofen, and Carter 2021).

Thus, the City of Warsaw created the Warsaw Food Lab (WFL), embedded in the organizational structure of the townhall and funded by The Food Trails Project. WFL became a platform for our design process. It brought together the representatives of the City of Warsaw (responsible primarily for formal aspects of the process and institutional setup), SWPS University (responsible primarily for evidence collection and analysis of collected data), and the team of professional facilitators (responsible primarily for substantive correctness of the process), supported by the involvement of external experts with sectoral expertise. The core design team comprised of eight people (three policy practitioners, two expert facilitators, and three researchers).

In general, WFL operates as an innovation incubator, focusing on creating and testing small-scale solutions related to food issues. Over time, the Warsaw Food Lab could become a support structure for the future food policy of Warsaw (the idea of Warsaw Food Strategy is currently under discussion).

3.2. Description of the timeline and content of the specification process

We documented the Food Trails project timeline using project records and participants’ memories. Below we provide an overview of the sequence of events in the project, while Table 1 sums up the detailed processes and decisions.

Table 1. Specification process and type of design activities in the Warsaw Food Lab.

#	Year	Months	Content of process and decisions	RE-DEFINEspecification	EXPLORE issue, situation	ANALYZE data, observations	CREATE solutions, interpretations	TEST insights, ideas, solutions
1	2021	Feb	The formal start of the food policy project with policy solution put upfront.	X
2	2021	Aug	The decision to focus on food waste problem.	X
3	2021	Sept–Dec	Exploratory study of food waste policy problem and mapping of existing food-waste initiatives in Warsaw.		X	X	X
4	2022	Jan	Series of workshops with stakeholders to test the ideas of different target groups for food waste problem in the context of the City of Warsaw.					X
5	2022	Feb	Decision to focus on 3 target groups: small restaurants, fresh fruits & veggies stands on local markets, households – students.	X
6	2022	Feb–May	Emergency response due to war in Ukraine followed by a change of target groups to (1) HORECA - small restaurants, and (2) Warsaw Food Bank (WFB).	X
7	2022	June	Exploration of food journey within the WFB system and a spectrum of stakeholders - the potential target groups of the intervention.		X	X	X
8	2022	July	Workshops with WFB management on the accuracy of insights gained during the exploration process.					X
9	2022	July	The decision to focus on selected stages of food journey and beneficiary organizations that cooperate with WFB.	X
10	2022	Aug–Sept	Analysis of the profiles of beneficiary organizations of WFB.		X	X
11	2022	Oct	The discussion with stakeholders on findings of the analysis - updating profiles of users and the perspective of the impact.				X	X
12	2022	Oct	The decision to focus on organizations cooperating with WFB all year and use donated food to provide meals to their users (food kitchens, homeless shelters, nursing homes).	X
13	2022–23	Nov–Feb	Systemic field study of targeted organizations (n = 57) - how they function, what their challenges are at every stage of the last mile of the food journey.		X	X
14	2023	March	Workshops with stakeholders on interpreting the results and setting a spectrum of desired policy outcomes.				X
15	2023	April	Checking with food experts the feasibility and rationality of stakeholders’ preferences for intervention outcomes.					X
16	2023	May	The final decision on the shape of the intervention and the expected effect.	X
17	2023	May–June	Workshops for detailed design of intervention architecture (instruments and causal pathway to effects) and field visits to selected target organizations.		X	X	X
18	2023	July	Final blueprint for intervention and check with stakeholders’ representatives.	X				X
19	2023	August	Final specification: Terms of Reference for contractor created.	X
20	2023–24	Sept–March	Pilot implementation and testing of the solutions in 2 organizations.					X

Launched in February 2021, Warsaw proposed the European project on developing sustainable food policy. The proposal already contained the idea for the solution - standard city market stall to sell local foods.

After six months of procedural work, The City of Warsaw set up the food lab structure with a core team of policy designers, which decided to address a more specific policy problem - tackling food waste. The primary rationale here was the risk of market disruption (the city introducing a public-funded veggie stand in a highly competitive market) and an increase in food prices, which prioritized food waste issues on the policy agenda.

Between September and December 2021, the design team conducted a literature review, expert interviews, and an overview of existing food waste initiatives in Poland to understand the scope of the food waste problem and map existing or future food-related initiatives led by Warsaw’s various departments. This lead to pinpointing three main food waste contributors - Households, Retail, and HORECA – and identifying six specific potential target groups: for HOERCA (1) small restaurants, (2) cafeterias and university canteens; for retail (3) small local stores, (4) veggie stands at local markets; for households (individual consumers) (5) students not living with their partents, (6) senior citizens.

The limited data on the degree to which specific groups contribute to the problem impeded evidence-based discussions. Hence, the focus was shifted to group attributes: group size, their significance or visibility to the city, and our existing knowledge about them. Synergies within the City of Warsaw also influenced decisions, particularly avoiding overlaps with other offices’ activities and focusing on target groups related to the office responsible for the project.

Once these considerations had whittled down the range of options, the conversation was driven by an assessment of feasibility, defined by the target groups’ receptiveness to change and potential challenges for intervention. Also, designers adjusted for initial over-optimism bias in planning, reducing the number of initiatives and target groups.

Early 2022 saw workshops refining these groups, finally choosing to focus on three target groups: (1) HORECA - small restaurants (5–15 tables, 30–60 seats), (2) Retail - fresh fruits and vegetable stands located at markets operated by the city, and (3) Households - Students living independently.

Russia’s invasion of Ukraine in February 2022 changed the situation. The subsequent influx of refugees to Warsaw highlighted food chain flaws, especially in food donation management. Post-crisis, the team kept HORECA – small restaurants, and changed to a new group: Warsaw Food Bank (WFB). In the following part of the text, we focus on the solution specification for the latter.

In June 2022, designers mapped the WFB food journey based on a literature review, interviews with WFB management, and review of internal WFB studies. They identified the following key actors in the WFB system: (1) donors (stores and corporations), (2) WFB staff, (3) beneficiary organizations, and (4) individuals consuming donated food. These activities culminated in a decision to focus on beneficiary organizations’ end-stage activities (compare: Figure 2). The design team profiled these organizations from August to September 2022 and decided to support those providing year-round meals to their users.

Figure 2. Food journey from donors to final users.

A comprehensive study into these organizations ran until February 2023, assessing their needs and constraints, leading to the formulation of three options for intervention’s outcomes: (a) help beneficiary organizations predict quality and quantity of received donations; (b) help beneficiary organizations prepare meals from "unpredictable" products; (c) help beneficiary organizations redistribute received food among themselves. Expert consultations in April 2023 emphasized the critical role of organizations’ entry points – unpacking, sorting, and storing donated food.

By May 2023, the team decided to focus on: (1) building routines among organization personnel for orderly unpacking, sorting, and storing donated food, and (2) building a routine among organization personnel for smart cooking from unpredictable products. May and June 2023 were devoted to developing details of the policy mix, thanks to workshops and field visits to the organizations-beneficiaries of the WFB.

The final blueprint of intervention (and therefore the product of specification) emerged in July 2023 (see Table 2). Based on this blueprint, the City of Warsaw prepared Terms of Reference and contracted out the pilot implementation (delivered at the turn of 2023 and 2024).

Table 2. Final theory of change developed for food waste intervention in Warsaw.

Target group	Action situations and obstacles	Intervention	Causal pathway and mechanisms	Effects
Focus on organizations-recipients of food bank, that use products to prepare meals for their users. Focus on the staff of those organizations - cooks and volunteers working in kitchen.	Focus on 3 types of products: fruits, veggies, and bread. Focus on three situations: unpacking & sorting, processing & storing, cooking. Focus on key obstacles: opportunity - limited and messy storage space, capability - high rotation of staff, capability - low know-how among kitchen volunteers.	Policy mix of equip, nudge, and boost: Equip with infrastructure – smart storage system, packaging machine, shredder; Equip with skills – self-guides to use infrastructure, on-boarding procedure, AI aid for cooking from unexpected products; Nudge & Boost for new routines – visual cues, room layout, decision trees, messenger effect (expert demonstrating innovations).	Reach: work with organizations that agree to test innovation. Actions: (a) storage system built, (b) staff trained and equipped with procedures, (c) expert introduced system. Obstacles removed: better storage, know-how codified, and easily available. Response: (a) staff uses new storage space, (b) staff uses AI for cooking, (c) staff uses new procedures and instructions.	Outcomes short-term(change of behaviors): (1a) staff developed new routines in the kitchen environment, (1b) staff feel more self-confident in the kitchen when dealing with donated food. Outcomes long-term(structural change): (2a) Higher % of donated food used (bread, fruits, veggies) in comparison to the situation before intervention; (2b) Food bank and other organizations declare willingness to implement the solution.
PREMISES on which decisions are built
This group of organizations is important for the city of Warsaw in the context of refugee adaptation. It is a relatively focused and small intervention – it is manageable and highly scalable in the case of success. Some good practices can be adapted from similar projects and routines in restaurants (already researched by the design team).	Three types of food are most often wasted, and at the same time, they are expensive and highly valuable for balanced diets. The chosen action situations determine the rest of the food journey in organizations. Change at the beginning of the process is more effective than work at the end of the food journey. Chosen obstacles are recognized by experts, food bank and organizations.	Shaping the environment reshapes behaviors. Thus, it is good to start by removing substantial barriers to opportunity. Know-how needs to be codified but also delivered easily. The staff has little time and attention to invest in learning new things. The system has to be self-explanatory because staff rotates.	Reach – organizations look for innovations that could provide them with savings. They also want to be leaders. Mechanisms driving response: Well-organized physical space makes desired behavior easier. Simplified and clear structuring of processes makes them easier to follow for people working under time pressure. Once people try a less burdensome process, they repeat it.	Change can be measured as part of the test – an experiment. There will be no substantial changes in the organizations during the experiment.

Based on: Project Food Trails – Warsaw, Poland.

3.3. Analysis of the dynamics of specification process

Our analysis started with looking at the dynamic of this process through the lenses of design activities performed in the lab (compare Figure 1). The Table 1 provides a summary of the analysis. Three things stand out. First, the process was iterative; there were several back-and-forth between re-defining specifications, exploring, and testing ideas. What started as idea of building veggies stands (#1 in Table 1), after literature review turned into idea of reducing food waste among fruits & veggies sellers (#5), then idea for supporting food bank system (#6) and eventually, after series of workshops and field research, ended with focus on food bank beneficiary organizations unpacking, storing and cooking routines (#12).

Second, creation is usually preceded by exploration and analysis. The segmentation in profiles of WFB’s users (#11) were developed with quantitative analysis of population of all bank users (#10), while policy outcomes (#14) were chosen after the in depth fields studies of chosen organizations (#13).

Third, almost every exploration and testing is followed by changes in specification that were further scoping the attention of designers. For example, exploration of food journeys from donors through food banks to users (#7 and #8) resulted in narrowing the scope to only organizations that use donated food regularly for preparing meals (#12). The action of getting food products and serving meals was further broken down into various organizational activities paired with a specific mix of interventions.

In summary, the analysis revealed that the design process under examination was iterative, exploratory, and adaptive. These repeated cycles of interactions refined design specification, progressively narrowing the focus to a specific stage of the food journey and the routines of food bank beneficiary organizations.

3.4. The synthesis of the outcome of the specification process

The final product of the all the activities of specification process is a Theory of Change presented in Table 2, and supplemented by the summary of premises on which decisions were made.

The intervention designed in the lab tackles food waste by focusing on organizations that receive donations from food banks to prepare meals. The intervention targets the staff - cooks and volunteers working in the kitchen. It concentrates on three main types of perishable products: fruits, vegetables, and bread, and three critical operational stages: unpacking and sorting, processing and storing, and cooking. It identifies major causes of waste generation, such as limited and disorganized storage space, the frequent turnover of staff, and the volunteers limited culinary skills. To overcome these challenges, the intervention employs a policy mix that includes equipping the organizations with a smart storage system, packaging machines, and shredders, enhancing skills through self-guides, onboarding procedures, and generative AI for cooking with unexpected product combinations. Furthermore, the intervention includes nudging and boosting the adoption of new routines through visual cues, optimized room layouts, decision trees, and the messenger effect, where an expert demonstrates the innovations.

The causal pathway of the intervention begins by partnering with organizations open to trying out these innovations. The action steps include constructing the new storage system, training the staff with procedures, and having an expert introduce the system. These actions aim to remove the identified obstacles by providing better storage solutions and making codified know-how readily available. The expected response is that staff will utilize the new storage, apply generative AI in their cooking, and follow the new procedures. Anticipated short-term effects are staff developing new routines and feeling more confident in the kitchen. Long-term outcomes are aimed at structural changes, including a higher percentage of donated food being used and a willingness among the food bank and other organizations to adopt the solution, thus scaling the intervention’s impact on reducing food waste.

4. Discussion of the case study findings

In this section we synthesize the observations from our case study, by addressing our research question: What does the specification of micro-level policy design look like regarding its merit and the process dynamics?

Regarding MERIT, we observed that design specification at the micro-level provides designers with a broader spectrum of choices than was initially conceptualized in the Introduction to this special issue. Designers address not three but six issues. They include (1) the designation of the target group, (2) expected outcomes (both immediate behavioral change and more systemic improvements), (3) time frame, (4) assumption of problem causes (obstacles to focus intervention on), (5) choice of instrument types, and (6) construction of a causal pathway with a change mechanism.

In addition, our case study clearly shows that some choices are discussed more often and in more detail, whereas other issues can be under-discussed or discussed briefly. In our case study, the choice of target group was revisited several times. With every change in the target group, a change in problem diagnosis followed. The ideas of the instruments (the whole policy mix) were contemplated at different stages of the designer’s work. At the same time, the expected outcome description remained mostly the same (to limit food waste), and only toward the end of the process did designers try to frame it in more precise terms of measurable behavior change. Relatively little time was given to consideration of mechanisms that could underlie the causal pathway. Thus, we must conclude that the most demanding part of the design, unpacking the black box of successful change, was underinvested in terms of time and attention.

Regarding the PROCESS DYNAMICS, we can conclude that the design specification was a highly iterative and interactive process. The REACT frame allowed us to trace this non-linear process. We observed that policy designers were continuously bouncing between two axes: working alone (explore and analyze) and interacting with stakeholders (create & test), simplifying reality into concepts (analyze & create), and confronting complex reality (explore & test). The multiple interactions with observable reality and stakeholders have been changing the content of the specification, which in turn put the process on different tracks.

We can conclude that designers have a relatively broad space to define the meaning and shape the final intervention, with substantial shifts on the ground.

5. Conclusions, recommendations and future directions

Based on our case study analysis, we make two main observations, followed by tentative suggestions for policy practitioners. The first significant finding is that micro-level design specification covers a spectrum of decisions related to the target group, action situations, obstacles, causal pathways with mechanisms, and effects. It provides quite a broad space for policy designers to reshape both understanding of the policy problem and change strategies to address it. The initially proposed conceptualization that distinguishes between specification (goals) and calibrations (tools) does not fit the reality of this case.

The second general observation is that micro-level policy design is complex and non-linear. Our analysis suggests that the specification is both engineering and bricolage. While the policy design process exhibited elements of systematic planning characteristic of an engineering approach, such as rigorous data gathering and analysis of policy problems, stakeholder mapping, and concept testing, it also involved a good deal of iterative learning, flexibility, and adaptation, illustrating a bricolage approach. This “mixed” approach was particularly beneficial when navigating the complex landscape of urban food waste, a problem space fraught with unpredictable dynamics and diverse actors. We propose the term "engineered bricolage" as a more accurate depiction of the dynamic nature of policy design specification.

Based on our case study analysis, we make the following recommendations. Concerning conceptualization, we propose a modification. The specification could cover the development of the Theory of Change, including the choice of instruments expected to trigger change and an explanation of causal mechanisms. Calibration could cover the plan to put those chosen tools and intended activities into operation (so-called Theory of Implementation - ToI). So basically, we conclude that specification is about developing the Theory of Change, while calibration is about ways to implement it. That adaptation would later help to distinguish between causes of potential policy/program failures: failure to put the intended policy tools into operation (implementation failure) vs. failure to trigger the desired effects with proper tools (theory failure). This distinction is well established in program evaluation literature (Scheirer 1987; Weiss 2007, 68) and practice (Vaessen, Lemire, and Befani 2020, 160)

Concerning design practice, we recommend using the Theory of Change as a canvas for the specification of the intervention. It has proved its utility in three situations:

1. When new stakeholders join the process (in our case, food bank management), ToC allows for explaining the nature of the problem and ideas for solutions and quickly integrating their ideas into evolving design specifications.

2. When designers communicate the intervention idea to outside actors (in our case, the leader, main funder (EC), and other cities in the consortium), ToC balances the intervention’s details and clarity.

3. When specification has to be turned into implementation (in our case, drafting Terms of Reference for the city’s contractors), ToC works as a blueprint that explains to implementers the goal and key elements that need to be in place to carry out positive change.

Summing up, ToC can be described as a “boundary object,” bridging the communication gap between different participants of the design process and providing a common frame to guide discussions and action (Stafford 2020, 9). We proposed a simple set of Theory of Change elements that can be useful for the practitioners conducting specification of intervention as well as academics who want to do comparative studies of interventions’ design (see Appendix).

We also acknowledge that the REACT frame allowed us to understand the changes in roles and situations during the design process, particularly switching between research and action perspectives and embedding into often messy reality vs. making more abstract, conceptual work. Thus, we recommend using the REACT frame for specific practical design assignments and comparative studies of design operations.

Notwithstanding these key findings, this study has some limitations. The main constraint lies in the single-case study design, restricting the generalizability of the findings to other contexts. While the in-depth exploration of one case provides rich insights, there is a need for comparative research that considers different contexts to test the robustness of our findings and the applicability of the "engineered bricolage" concept in different policy domains.

Our case was specific in terms of organizational setting. Policy labs are a relatively new phenomenon in public administration practice. They are an "odd addition" to traditional bureaucracy. Labs have relatively little political influence (focus on merit, often smaller scale policies), flexible methods that allow experimentation, and often, as in our case, secured external and project-based funding with relatively low time pressure.

We also suggest further examining the role and political dynamics of different actors. Our study focused on policy designers, assisting researchers and highly determined stakeholders. Examination of a broader spectrum of stakeholders is needed.

In sum, this study contributes to a nuanced understanding of policy design’s “last mile.” It shows the process of turning policy objectives into concrete actions, underlying the need for an adaptive, creative, and inclusive approach that combines systematic planning with flexibility and on-the-ground learning. This work could be a potential stepping stone for future research exploring the complexities of policy design in the face of evolving societal and environmental challenges.

Acknowledgments

The authors would like to express their deepest gratitude to colleagues from the team of the Warsaw Food Lab who developed policy intervention described in the case study. The authors would like to thank: Natalia Boitot, Kamila Czerwińska, Paulina Grabowska, Michał Kuszewski, Bartosz Ledzion and Katarzyna Wrońska.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and publication of this article: This work was supported by the National Science Center, Poland, grant number 2021/43/B/HS5/01935.

Appendix:

Template for design specification

This appendix presents our proposal for the canvas for design specification. It is built on streams of literature from public policy (Breuer et al. 2016; Davison et al. 2020; Maru et al. 2018; Olejniczak, Śliwowski, et al. 2020), program evaluation (Weiss 2007; Rogers and Funnell 2011; Mayne 2017) and behavioral insights (Bartholomew and Mullen 2011; Michie, Atkins, and West 2014; Davison et al. 2020; Kettle and Persian 2022), and was tested in the current case study.

Canvas is visualized in Table 3. Each column represents one element of the Theory of Change (ToC), and it contains short questions to guide the discussions of the participants of the design specification.

Table 3. Proposed template for designers – specification of theory of change.

Target group	Action situations and obstacles	Intervention	Causal pathway and mechanisms	Effects
CHOICES to be made for design specification
Choice of Target Group What Target Group is relevant to policy issue?	Choice of action situation and obstacles What are the key ACTION SITUATIONS of the target group? What are the major OBSTACLES on the journey of Target Group regarding their capability, opportunity and/or motivation?	Choice of a mix of instruments What INSTRUMENTS will address the needs of Target Group in terms of capability, opportunity and/or motivation?	Choice of steps in causal pathway and driving mechanism How will the change of behavior progress? In particular: How will Target Group be reached with intervention? What actions will be performed for Target Group? What will Target Group gain in terms of capability, opportunity and/or motivation? How will Target Group behave after obstacles removal?	Choice of indicators describing effects How success will look like. In particular: What are indicators of the new, desired behavior of the Target Group? What are indicators of systemic benefit coming from this behavior change?
PREMISES to be articulated
Why this Target Group was chosen?	Why this action situation and obstacles were chosen as the focus of the intervention?	Why have those instruments been chosen?	What mechanisms will drive the response of the Target Group?	How could the actual change be captured (causation vs correlation)?

The canvas starts with an approach centered on the actor and his behavior [Column: Target Group]. It links policy issues with specific target groups whose behaviors seem instrumental to the positive change (Koleros et al. 2018).

Once the actor and its behavior are chosen, designers focus on mapping in what situations behavior can occur and its driving and restraining forces [Column: Action Situations and Obstacles]. The logic of our ToC follows the Lewinian proposal for inducing change that favors reducing the "restraining forces" over increasing the "driving forces." (Crosby 2021; Kahneman 2013). Thus, we focus on identifying the main obstructions to behavior.

For identifying situations where restraining forces hinder the target group, designers can use the concept of action situations (Ostrom 2005) – a series of activities and decisions the target group performs. Design literature calls this "decision journey" in user’s interaction with products and services (Kumar 2012).

For unpacking specific behaviors and naming restraining forces, designers can use the typology of obstacles: deficits in capability, barriers in opportunity, and adverse motivations of the target group. It is based on a well-established COM-B model, explaining that behavior is an interplay between capability (knowledge, skills, and physical capacity to perform behavior), opportunity (factors in the physical or social environment influencing behavior), and motivation (reflective and automatic mechanisms that energize or inhibit a behavior) (Michie, van Stralen, and West 2011).

When it comes to inducing change and removing restraining forces [Column: Intervention], designers can accommodate the typology of policy instruments - a combination of equip, ban, dis/incentivize, inform, boost, and/or nudge to help their target group (Olejniczak, Śliwowski, et al. 2020).

The change is portrayed as a causal chain with underlying mechanisms [Column: Causal Pathway and Mechanisms]. This is aligned with evaluation practice (Mayne 2017; Lemire et al. 2020) and first-order mechanisms in public policy literature (Capano, Howlett, and Ramesh 2019).

The final column on the canvas is devoted to effects, which we understand as short-term changes in the behavior of the specific target group and a longer-term systemic change. This distinction between types of effects is well recognized in the evaluation literature (Newcomer, Hatry, and Wholey 2015).

The table also provides the bottom row with premises for each element of ToC. That row allows articulating reasons for which decisions are made. This should also increase the clarity of change mechanisms assumed in causal pathways – a challenge often pointed out in the Theory of Change literature.