3.1 Introduction

This chapter outlines a practical framework for designing scalable technology solutions that solve development challenges. We begin with an overview of the common constraints to sustainable development that often are encountered in the context of poverty. These constraints are based on a large body of research in development economics, political economy, psychology, and other social sciences; and they help to explain why engineering innovations so frequently fail to achieve outcomes when implemented in the real world. In the second part of this chapter, we provide a framework for implementing development engineering projects, consisting of four key activities: innovation, implementation, evaluation, and adaptation. Combining these activities in an iterative (and usually nonlinear) path allows the researcher to anticipate and design around the most common pitfalls associated with “technology for development.”

3.2 Innovation Under Constraints

To find solutions to thorny development challenges, researchers must begin with building a deep understanding of context and environment. To some extent, this can come from direct observation—from researchers embedding themselves within representative communities, observing the cadence of daily life, learning how it is to walk in the shoes of the potential users of a future innovation. This approach is central to the success of product design firms like IDEO (Kelley, 2005).

Until recently, direct observation (and other elements of human-centered design) has remained relatively uncommon in the technocentric approaches to “engineering for development” found in many elite universities. Yet development economists, political scientists, and others in the social sciences have invested decades in such work, and in the collection of descriptive data across populations and contexts. This has resulted in generalizable findings about the market systems, institutions, behaviors, and social norms governing life in many low-resource settings. Learning to systematically apply these insights to the design of a novel technology is an essential thrust of development engineering.

This section provides an overview of the common constraints encountered in many developing countries—and to some extent in low-resource communities throughout the world. Without judgement, these constraints are actually just alternatives to the “ideal” market systems and institutions imagined to exist in wealthy countries. In some cases, they have emerged as critical adaptations to local conditions (such as resource scarcity, conflict, colonization, and ethnic diversity). In the engineer’s mindset, we can think of these conditions as design requirements, because they can affect the adoption, performance, impact, and scaling of a technological solution. We can also think of these constraints, themselves, as targets for intervention (Soss et al., 2011). For example, predatory policing, which may be observed as a constraint to economic development, could be directly targeted through the design of mobile applications and political reforms that empower citizens to monitor and report police activity.

In a sense, the most basic constraint faced by people living in poverty is income uncertainty. For survival, humans require continuous access to food, water, heating, cooling, and shelter. Yet poor households, by definition, experience scarcity—not just lack of income but also income that is lumpy across time (Collins et al., 2009). This makes it difficult to invest in basic needs, let alone new technologies or other assets. In urban settings, this lumpiness may take the form of irregular income from small family-owned enterprises. These businesses are often constrained by a lack of formal access to capital (in the form of savings or credit products). As a result they find it difficult to invest in the inventory, marketing, supply chain tools, and other inputs needed to build more reliable profits.

It is a more complex story for households reliant on farming for survival. Agriculture employs the majority of the world’s poor, typically on small family-owned farms. Income from agriculture is seasonal by nature: profits are generated largely at harvest time. This cyclic pattern of production creates lumpiness in household consumption (Mobarak and Reimao, 2020). In addition, productivity is dependent on weather and climate conditions, which are highly unpredictable and can vary substantially from season to season or from year to year. This uncertainty makes it difficult for households to purchase goods or services on a regular basis, and it can also deter households making investments in new technologies (even when the longer-term economic benefits of a technology are well understood).

Beyond lumpy consumption, households face a lack of access to savings, credit, and insurance products, all of which are useful for managing risk and smoothing household consumption. This unmet need for financial services—combined with unpredictable “shocks” like climate change, illness, and death—means that many poor households are risk averse when it comes to spending on new products and services.

There are a host of other constraints encountered by low-income households in developing countries (and in many developed countries). In this chapter, we will outline three classes of constraints that the development engineer should consider: (1) market constraints; (2) institutional constraints; and (3) behavioral and social constraints. You may not encounter all of these constraints in a given project, and those encountered may not be binding (meaning that they may not be the bottleneck we need to target). However, they are useful as diagnostic and design tools (Hausmann et al., 2008), and they can help explain why technologies that have worked in “developed” settings may fail when transplanted to a new setting.

These constraints can be considered as design requirements, but they can also be direct targets for technological innovation. Where markets fail to meet the needs of poor households, there may be a technology—say, the capture of real-time information on prices—that can level the playing field for disadvantaged consumers. When institutions have been captured by elites (creating conditions for inequality), there may be innovations that enable the decentralization of assets or force transactions to be more transparent to citizens. While some of these problems require policy reform, others may be amenable to technological intervention.

3.2.1 Market Constraints

Markets are the mechanisms through which goods and services are produced, distributed, and consumed; and well-functioning markets can generate clear signals of supply and demand, partially transmitted in the form of prices. In reality, all markets operate imperfectly, and every country suffers from market distortions (or “failures”) that result in the inefficient allocation of resources. Yet the developing world is particularly complex.

In most developing countries, the economy is dominated by the informal sector, which consists of market activity that is not organized, monitored, or regulated by government. This informality, combined with challenges like weak infrastructure and high transport costs, inhibits the development of modern, market-based economies. Informality also reduces government tax revenue and the state’s ability to redistribute resources through public benefits programs. As a result, markets in developing countries often fail to efficiently allocate the supply of goods and services to those with greatest demand.

Informality may enhance resilience in some communities and contexts; however, it also intensifies the uncertainty that poor households already deal with. Understanding informal and imperfect markets—and anticipating their effect on the performance and sustainability of a technology—is key to designing a product or service that will achieve development impacts. A brief summary of common market constraints is outlined in Table 3.1.

Table 3.1 Examples of market constraints
Full size table

3.2.2 Institutional Failures

Organizations, in particular government bureaucracies and nongovernmental organizations (NGOs), play a critical role in delivering basic services to people in developing countries—from water, sanitation, and education to pensions and social protection schemes. Many low-income households rely on these formal institutions, whose operations are guided by written rules and laws, for their welfare. At the same time, people in resource constrained settings also rely on informal arrangementsFootnote 1, like social networks, based on kinship or caste for accessing services. For example, it is common for villagers in rural settings to finance loans or emergency support from family members or money lenders within the village.

Both formal and informal institutions can introduce inefficiencies and distortions in implementation of new technologies or policies (Helmke & Levitsky, 2006). For example, ethnic or provincial community leaders may hold socially important positions in communities, limiting the power of government appointees to manage local affairs. In the absence of effective community oversight, these local leaders can control the functioning of the state apparatus and capture public goods for private benefits (a process known as “elite capture”; see Bardhan & Mookherjee, 2000). Governments may find that their policies fail to achieve outcomes for disadvantaged communities, or that outcomes differ from a policy’s stated objectives. For example, states in resource constrained settings tend to generate less tax revenue than targeted, due to weak collections and audit capacity as well as missing infrastructure (Acemoglu et al., 2011).

In all parts of the world where formal institutions are inefficient or weak, informal institutions remain relevant and effective at meeting the needs of citizens. Indeed informal institutions, like their more codified counterparts, can establish and enforce rules, negotiate disputes, distribute shared resources, and constrain social behavior. However, informality is also challenging for the scale-up of a technology: Informal institutions often follow tacit rules, known only to “insiders.” By their very nature, informal norms and institutions (particularly those without written record) require context-specific understanding. Researchers who want to successfully implement and scale up new technologies need to invest time and resources in trying to understand how informal institutions behave. A few examples of commonly encountered constraints are given in Table 3.2.

Table 3.2 Examples of institutional constraints
Full size table

3.2.3 Social Norms and Behaviors That Constrain Development

Communities and individuals living in poverty face unique behavioral and cognitive constraints that affect their decision-making about technology. Further, social norms at the community and the household level also have the potential to shape decision-making around technology adoption in resource constrained settings. Some of these are outlined in Table 3.3.

Table 3.3 Examples of norms and behavioral constraints
Full size table

Reviews of Market, Institutional, and Behavioral Constraints

Below is a collection of practical white papers outlining the constraints faced by households, institutions, and markets in developing economies. These are organized by sector, making them a useful resource for engineers and development practitioners. Additional reviews of the evidence from international development research can be found at the International Initiative for Impact Evaluation (http://3ieimpact.org).

  • Agriculture

Experimental Insights on the Constraints to Agricultural Technology Adoption (2019), accessed at https://escholarship.org/uc/item/79w3t4ds

Market inefficiencies and the adoption of agricultural technologies in developing countries (2013), accessed at https://escholarship.org/uc/item/6m25r19c

  • Governance

Governance Initiative Review Paper, J-PAL Governance Initiative (2019). J-PAL Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/review-paper/GI_review-paper_2019.pdf

  • Digital Identities

Digital Identification & Finance Initiative Africa: An Overview of Research Opportunities (2019). J-PAL Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/review-paper/DigiFI_framing-paper_june-2019.pdf

  • Post-Primary Education

Expanding access and increasing student learning in post-primary education in developing countries: A review of the evidence (2013). J-PAL Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/2020-03/PPE_review-paper_executive-summary_2013.05.07.pdf

  • Gender and Women

What Works to Enhance Women’s Agency: Cross-Cutting Lessons from Experimental and Quasi-Experimental Studies (2020). J-PAL Working Paper, accessed at https://www.povertyactionlab.org/page/what-works-enhance-womens-agency

  • Labor Markets

Reducing search barriers for job seekers (2018). J-PAL Policy Insights, accessed at https://doi.org/10.31485/pi.2234.2018

Urban Services Improving Access to Urban Services for the Poor: Open Issues and a Framework for a Future Research Agenda (2012). J-PAL Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/2020-03/USI_review-paper.pdf

  • Youth and Employment

J-PAL Youth Initiative Review Paper (2013). Abdul Latif Jameel Poverty Action Lab Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/documents/ YouthReviewPaper_March_2013_0.pdf

J-PAL Skills for Youth Program Review Paper (2017). Abdul Latif Jameel Poverty Action Lab Working Paper, accessed at https://www.povertyactionlab.org/sites/default/files/review-paper/SYP_review-paper_2017.pdf

3.3 Framework for Research

As the previous section outlines, the success of any innovation requires deep understanding of the constraints—market, institutional, and behavioral—that can prevent adoption of a development solution and its impact at scale. These constraints inform the design of technology, and they can also affect our ability to implement high quality research. Years of field work have resulted in a set of best practices that we present here, as a practical framework for advancing promising technologies from the lab to the field. The most important guiding principle behind this framework is its emphasis on feedback and iteration and the avoidance of a linear implementation process.

The practice of development engineering focuses on the entire arc of innovation—from problem discovery and technological invention, to prototyping and pilot testing, impact evaluation, and finally adaptation for scale-up. These stages are part of a continuum, and they are not necessarily carried out in sequence. In some cases, the real-world evaluation of an existing product will lead to the design of an entirely new technology, based on iterative feedback from users. The chlorine water dispensers developed by Miguel, Kremer, and colleagues is one example (Null et al., 2012). In other cases, a novel technology will enable measurement of development outcomes at higher frequency or resolution, leading to the discovery of new problems and opportunities (Blumenstock et al., 2016; also see Chap. 15 in this textbook by Wilson).

Fig. 3.1
figure 1

Iterative framework for development engineering

Full size image

Each of the following chapters in this textbook will describe a unique research workflow, but they all fit within the framework of four activities: innovation, implementation, impact evaluation, and adaptation for scale (see Fig. 3.1).

Innovation

Innovation is at the heart of every development engineering intervention or solution. It is the process of discovering and characterizing a problem and then developing a generalizable technological solution—one that can address the challenge at scale. The innovation can lie in adapting technology to solve a new problem (e.g., by bundling an emerging or existing technology with a novel economic, political, or behavioral intervention); or it might lie in designing an entirely new technology around any of the constraints outlined earlier in this chapter. The innovation may even lie in creating new ways to measure development outcomes, either through instrument design (like a wireless cookstove sensor) or the design of new analytic techniques (like the use of remotely sensed imagery to predict household asset wealth).

However, the discovery of a suitable problem, and the development of the design requirements for a solution, is never linear; it requires a critical and evolving understanding of local context. As an example, we consider the design of a treatment system for removal of arsenic from drinking water. First, we investigate the experiences and environment of households affected by arsenic contamination (mostly people living in rural Bangladesh and Eastern India). In this context, shallow tube wells are the main source of drinking water, household asset wealth and consumption are low, and willingness to pay for Arsenic-safe water is lower than willingness to pay for piped water (Khan et al., 2014). Perhaps there are collective action failures in the maintenance of existing water infrastructure, and community trust of outside commercial providers is limited (Alfredo and O’Garra, 2020).

The engineer’s goal is to solve the problem of access to clean water, using a combination of technological and socio-economic innovations to overcome these hypothesized constraints. The solution must address market failures, institutional challenges, and the preferences and behaviors of people facing arsenic poisoning—in addition to their public health needs. These constraints can be shaped into the solution’s design space, in the form of performance parameters (e.g., failure tolerance, reliability, salience, desirability, cost, accessibility). It is also important to anticipate any negative externalities created by the innovation, such as wastewater production and environmental contamination.

Once the problem has been defined, the constraints characterized, and a solution posed and prototyped (often just on paper), the researcher can begin to articulate a theory of change: a set of hypotheses about how the proposed solution overcomes observed constraints. This is the key output of the innovation activity: prototype brainstorming and a set of hypotheses that are to be tested. But how do we get here?

To develop basic insights about a community’s development challenges, researchers often use qualitative approaches like ethnographic observation and human-centered design (HCD).Footnote 2 Much has been written about these methods, and we refer readers to a few resources in Box 3.1. In addition to these methods, development engineers often consult existing data and survey research to understand their targeted communities. Well-designed surveys can offer a quantitative and representative view of users’ perceptions and preferences. Examples include nationally representative datasets like the Living Standards Measurement Surveys (Grosh & Glewwe, 1995) and the Demographic and Health Surveys (Corsi et al., 2012), as well as survey research projects published in journals of development economics and political economy. In addition, national statistical offices in most countries publish census data at intervals, and they may release other de-identified administrative datasets. Large nongovernmental organizations (NGOs) often publish their own datasets, in areas like education (Banerji et al. 2013; Mugo et al., 2015), health (Murray et al., 2020), and politics (Afrobarometer Data). Resources like these are described and referenced in many of the chapters in this book.

Of course there are many challenges in collecting reliable survey data in low-resource settings (Iarossi, 2006), and these have been deeply documented across a range of domains—from health, nutrition, and gender to welfare and politics (Caeyers et al., 2012; Glennerster et al., 2018; Lupu & Michelitch, 2018). Issues with survey data include sampling errors, respondent biases, and small sample sizes (because budgets for survey research are always limited). There are other sources of unexplained variation in survey data, as well—artifacts that are introduced throughout the surveying process. These range from the selection of interviewers (West & Blom, 2017) and nuances in the wording or order of survey questions (Blair et al., 2020) to the length of a survey and how you compensate study participants (de Weerdt et al., 2020).

To improve the reliability and reproducibility of survey results, some researchers carry out intensive qualitative research with random samples of individuals surveyed in quantitative surveys, particularly in cases where survey questions address sensitive issues like corruption, crime, or other risky behaviors (Blattman et al., 2016). Some researchers now publish their survey protocols or use published questionnaires that have been validated against more reliable methods of data collection (Meyer et al., 2015). Still, there are few repositories that allow you to browse and search for questionnaires by geography, population, or topic. You will often need to sift through the supplementary materials published as part of academic journal articles (in fields as diverse as sociology, anthropology, economics, political science, and public health) to discover existing survey instruments.

To supplement survey datasets, researchers increasingly leverage digital technologies that capture complementary and (in some cases) less subjective information about context within their communities of interest. For example, financial transactions using mobile phones or debit cards can offer a view into consumer behavior including purchasing patterns (Bachas et al., 2017), loan repayments (Björkegren & Grissen, 2018), and social insurance mechanisms (Blumenstock et al., 2016). Remotely sensed data—like satellite imagery or drone video footage—allow us to directly observe agricultural yields and management practices from the sky (Lobell et al., 2020). Data extracted from social media platforms can expose relative poverty (Fatehkia et al., 2020) as well as popular sentiment and prevailing social norms, using natural language processing to automate analysis (Calderon et al., 2015). Technology companies like Facebook have begun leveraging internal, georeferenced user interaction logs to produce a range of datasets, from human population density to international firm surveys (Stevens et al., 2019; Schneider, 2020). Anonymized call detail records and geolocation data from mobile phones can also reveal household outcomes—from consumption patterns and asset wealth to migration decisions and response to violence (Aiken et al., 2020; Blumenstock et al., 2015; Chi et al., 2020; Blumenstock et al., 2018).

Insights obtained using these large-scale datasets can be useful to understand context and also to measure development outcomes. However, few of these methods have been extensively validated against “ground truth.” In addition, these digital data sets carry their own biases, for example, based on who has access to mobile technology or on which communities have been surveyed enough to train a machine learning algorithm based on satellite imagery.

Ultimately, the data take us only so far. It takes decades to build deep knowledge of the development constraints facing any country or community. It requires knowledge of domestic and regional politics and economic history; it requires familiarity with local views about colonialism and its legacies. It also requires understanding a nation’s struggles with ethnic and gender identity. Perhaps this creates a natural imperative to collaborate with researchers, policymakers, and civil society organizations based in the communities you wish to empower. Inclusive, respectful partnerships with local actors are key to many successful development engineering projects, and this success relies on the alignment of incentives for all participants.

Resources for Defining Development Challenges

  • Qualitative and Design Methods

    • Field work for development research (Scheyvens, 2014)

    • Qualitative research methods (Marshall & Rossman, 2014)

    • Participatory action research (Kindon et al., 2007)

    • Human-centered design methods (Holeman & Kane, 2020)

  • Quantitative survey research

    • Academic research surveys of specific populations, including in-person enumerated surveys or mobile surveys (Rossi et al., 2013)

    • Nationally representative surveys and census data (Nikolov, 2009)

    • International surveys like the demographic and health surveys (Corsi et al., 2012) and Living Standards Measurement Study (Grosh & Glewwe, 1995)

  • Administrative Data

    • Transactions records generated in the delivery of services by health systems, schools, government agencies, cooperatives, and other public organizations (Meyer & Mittag, 2019)

    • Customer transaction records generated by retailers and other private sector firms (Di Clemente et al., 2018)

  • “Big” Data

    • Social media traffic scraped from public sources or accessed through agreements with technology firms (Calderon et al., 2015)

    • Satellite imagery, including public and private assets (Jean et al., 2016)

    • Anonymized call detail records accessed through agreements with mobile network operators (Blumenstock et al., 2015)

    • Geolocation data captured through consumer smartphone applications (Williams et al., 2015)

    • Networked sensors (e.g., personal activity monitors, grid electricity sensors, and precision agriculture devices) that capture large volumes of environmental or behavioral data (Ramanathan et al., 2017)

Implementation

Implementation is the process of piloting a new innovation, monitoring its technical performance in the field, and understanding the factors that influence “effective” implementation. This is in sharp contrast with the linear (and often idealized) model of the engineer moving directly from invention to impact. An iterative approach is required—a multistage progression from invention, to pilot, to full deployment—with feedback loops. We must frequently return to earlier stages of our research to update our hypotheses and theories of change. Each assumption is revisited in light of the data and insights gained from the previous iteration. This iteration—the process of advancing, updating, and retrialing—can start within the lab. But it is also part of the move from the lab to field trials or the expansion from one market to another.

In industrialized countries, such iterative loops are relatively commonplace in product development. Technology firms have specialized teams focused on product marketing, user interaction, design, product management, engineering, quality control, sales and growth, and financing. Because these activities are well-resourced and have been well-studied, they result in relatively reliable processes of iteration. In the context of development engineering, often the same team is blessed with the burden of taking a technological invention from the lab, all the way to product development, evaluation, and distribution. There are plenty of stumbles and failures along the way. This framework therefore focuses on training the practitioner to cultivate a learning mindset, treating each iteration as a valuable learning opportunity, and remaining ready to investigate and pivot when outcomes diverge from expectation.

The implementation stage does not focus on technological prototypes alone. It also involves the design, testing, and refinement of different business models (or delivery models) for a technological solution. At this stage, it is useful to test hypotheses about users’ willingness to pay or about how they access information. This is a good time for using experimental methods that reveal the demand for a new product or service, including pricing experiments as well as behavioral games (like take-it-or-leave-it studies) that can reveal people’s preferences (Dupas et al., 2013).

At this stage, the development of sound partnerships also becomes paramount. Innovating in a resource-constrained setting can be challenging because it often requires coordination across multiple partners, each with differing standards, norms, and incentives. One partner may support small-scale manufacturing, while another carries out field testing. Still other partners may be needed to collect user feedback or implement rigorous evaluations. The researcher is often reliant on local partners to understand the local context and effectively implement studies, deployments, and experiments. These partners may deprioritize the project or deviate from agreed plans, because of internal challenges or as a response to the external environment.

The weakness of government or community institutions can also constrain the ability to implement a project effectively. Piloting of new technologies may be regulated by governments, and research involving humans is always overseen by local review boards. Yet a lack of transparency can make it challenging to obtain the necessary permissions for experimentation. Researchers must learn local processes and find ways to overcome institutional challenges. Iterative implementation allows the researcher to discover the optimal implementation strategy over time, learning how the solution will ultimately perform in the targeted setting.

Evaluation

The evaluation component of the development engineering framework focuses on using scientific approaches (i.e., randomized controlled trials and quasi-experimental methods) to isolate the causal impacts of an innovation. We are interested in understanding the effects of a new solution on human and economic development; we also want to accumulate knowledge that can generalize to new contexts. Evaluation plays a central role in development engineering, in part because the field is in strong need of better evidence. If we fail to learn from our work, we will perpetuate the “valley of death” between tech-for-good innovations and their successful scale-up.

In this framework, we emphasize evaluations that test hypotheses and rigorously investigate how a technology affects health, economic, and other outcomes in the “real world.” We are also interested in exposing barriers to technology adoption, and testing our theoretical models about the optimal delivery of a technology. In some cases, we may want to measure spillovers (or unintended consequences) of an innovation or to assess long-term effects. While a short-term evaluation might show strong initial take-up, later follow-ups often expose disuse, environmental costs, and other failures that dampen benefits.

We may use surveys to collect self-reported outcomes, or we may use sensors and other networked devices to automate the monitoring of outcomes. We may choose to instrument our solution, so that ongoing evaluation is incorporated into operations. Regardless, by designing your evaluation carefully, you can investigate whether the failure of a technological solution stems from technology design itself or from the delivery model. If the failure is due to a flawed business model, evaluation results can be used to modify pricing or design a new financing scheme. If the failure has to do with technology design—for example, if the volume of human waste collected from a community is too limited to support continuous urea extraction—the results of evaluation can be used to refine design parameters and develop a more appropriate solution. A good evaluation allows for design iteration—with researchers incorporating the feedback into redesign—and also yields generalizable knowledge that can be applied in new contexts.

While many researchers will evaluate their solution at pilot scale, with careful control over implementation, there can also be value in deploying and evaluating a technology at large scale (e.g., nationwide). This teaches us something about the effectiveness of a solution at scale, when it is implemented under less controlled conditions (see Chap. 20, on evaluation of Aadhaar, India’s national digital authentication program). There is also value in evaluating a solution deployed across multiple contexts, in tandem, through portfolios of field experiments that test a shared hypothesis. This teaches us about the variations in effectiveness across conditions and can also reveal the sorts of adaptations that are required for an innovation to succeed at large scale (ref meta-keta, WSH eval).

Adaptation

Scaling up involves taking an innovation from the evaluation stage (with evidence of positive impact, albeit on a limited number of users) and adapting it to reach a larger number of users and to reach users in new geographies. The process of maintaining a technological solution at scale comes up with its own unique challenges. For example, for scale-ups that rely on market processes, the business model becomes a critical factor determining long-term sustainability. There will be challenges in managing deep supply chains, which requires strategies for mitigating risks from the market frictions commonly found in developing countries. Operations and maintenance, along with monitoring for quality assurance, will require critical attention. In addition, customer “success” may require not only technical support but also costly investments in user training or onboarding, particularly for communities that have not interacted extensively with your class of technologies. Some innovations will require intellectual property (IP) protections to ensure broader use and scale-up. However, IP regimes in resource poor countries may be poorly designed or weakly implemented, making IP protection a risky choice for researchers.

For scale-ups implemented in partnership with governments, it is key to navigate the political economy of institutions (and the incentives of those with vested interests) as well government regulation, legal challenges, and the role of civil society in oversight. Further, public institutions responsible for implementing or disseminating services may fail to adhere to the researcher’s well-defined standards, which can compromise the fidelity of implementation. This is an issue in many under-resourced communities, where there are high rates of absenteeism among frontline workers who are also overburdened with administrative tasks and responsibilities (Finan et al., 2017). The prevalence of corruption, along with weak monitoring of government workers, can also make implementation of projects a challenge.

The success of scale-up efforts is closely linked to the concepts of evaluation and iterative implementation. Evaluations conducted as part of small-scale field pilots allow the researcher to understand the challenges of implementation and gather evidence of a solution’s impact. Ideally these evaluations also reveal the mechanisms through which a product acts and expose any required or enabling conditions. This generalizable knowledge enables scale. As we move from the pilot context to a larger scale or from one country to another, we can then test whether the conditions for intervention success are found in new target environments and target scaling efforts where the innovation is most likely to achieve impact (Bates & Glennerster 2017).

Several successful examples of scale-up, from chlorine dispensers to provide clean water to deworming tablets , were first piloted at a smaller scale. There are also examples where attempted scale-up without evaluation led to failure. For example, the Embrace infant warmer developed to work in poor countries with limited healthcare facilities (Pg 71, Jugaad) proved effective in pilots but failed to find traction after initial adoption.

The case studies that follow are written to tie these processes—innovation, iterative implementation, evaluation, and adaptation—together. They demonstrate how feedback from one stage informs the next. The framework (and this textbook) will also undergo iteration, as new ideas are incorporated over time (Fig. 3.2).

Fig. 3.2
figure 2

The path of development engineering

Full size image

3.4 Additional Resources

In addition to this textbook, there is an expanding pool of resources available to researchers in Dev Eng. These include the open access peer-reviewed journal Development Engineering: the Journal of Engineering in Economic Development. This journal publishes original research across multiple areas of Dev Eng, including:

  • Engineering research and innovations that respond to the unique constraints imposed by poverty

  • Assessment of pro-poor technology solutions, including field performance, consumer adoption, and end-user impacts

  • Novel technologies or tools for measuring behavioral, economic, and social outcomes in low-resource settings

  • Lessons from the field, especially null results from field trials and technical failure analyses

  • Rigorous analysis of existing development “solutions” through an engineering or economic lens