Working paper 589

Navigating adaptive approaches

for development programmes
A guide for the uncertain
Jamie Pett
September 2020

• Adaptive approaches have emerged in several sectors, including software development, product
and service design, technology startups and international development.
• Adaptive approaches can help practitioners counteract misplaced certainty. By talking to potential
Key messages
users, understanding institutions, interests and ideas and investigating the root causes of a
problem, practitioners applying these approaches can illuminate the underlying nature of the
problem and context.
• Rather than building a whole solution straight away, these approaches commonly encourage
practitioners to start small and use structured cycles of testing and learning. There is scope to
further consider how different approaches can be better brought together and combined.
• Adaptive approaches in development provide a wider range of options for what to create and
facilitate – not only products or services, but also forms of collective action. There are also
alternative ways to think about scale – considering how others might take up an idea and looking
for leverage, rather than quantity.
This material has been funded by UK aid
from theThis material
UK Government, hasthebeen
however views funded by UK aid from the UK Government, however the views expressed do not necessarily reflect the UK Government’s
expressed do not necessarily reflect the UK
official policies.
Government’s official policies.

Readers are encouraged to reproduce material for their own publications, as long as they are not being sold commercially. ODI requests due
acknowledgement and a copy of the publication. For online use, we ask readers to link to the original resource on the ODI website. The views
presented in this paper are those of the author(s) and do not necessarily represent the views of ODI or our partners.

This work is licensed under CC BY-NC-ND 4.0.

Acknowledgements

Many thanks to Richard Butterworth, Ben Kumpf and other Foreign, Commonwealth and
Development Office (FCDO) staff who provided comments on drafts of this working paper, and Lea
Simpson and Abigail Freeman from Brink. The author is grateful to Craig Valters, Leni Wild and Alina
Rocha Menocal for their detailed comments and support. Thanks are also due to Ben Ramalingam for
acting as peer reviewer, and to Matthew Foley for copyediting and Paola Abis for production support.
Responsibility for the content of this paper, as well as any errors and omissions, lies with the author.
This working paper has been produced as part of the FCDO’s LearnAdapt programme, a
collaboration between FCDO, ODI, Brink and Feedback Labs to explore how to manage adaptive
development programmes better.

About the author

Jamie Pett is a facilitator and researcher who has worked on LearnAdapt as a contractor since 2017.
In this role, he helps the FCDO to create an enabling environment for adaptive and innovative
programmes. His interests include systems change, developmental evaluation, adaptive management
and changing power dynamics in the aid sector.

3
Contents

Acknowledgements3

List of figures and tables 5

Acronyms6

Executive summary 7

1 Introduction 8

2  Introducing each approach 10

2.1 Agile 10
2.2  Human-centred design 11
2.3  Lean startup 12
2.4  Thinking and working politically 12
2.5  Adaptive management 13
2.6  Problem-driven iterative adaptation 13
2.7  Common principles of adaptive approaches 14

3  Adaptive approaches throughout the programme cycle 15

3.1  Enabling adaptation 15
3.2  Assess and design 16
3.3  Implement and adapt 18
3.4  Adapt with the end in mind 22

4 Conclusion 24

References25

Annex 1  Summary of adaptive approaches 29

4
List of figures and tables

Figures

Figure 1  The UK Design Council double diamond 11

Figure 2  Adaptive approaches through the programme cycle 16

Figure 3  Think like a system, act like an entrepreneur 18

Tables

Table A1  Summary of adaptive approaches 29

5
Acronyms

DFAT Department of Foreign Affairs and Trade (Australia)

DFID Department for International Development (UK)
FCDO Foreign, Commonwealth and Development Office (UK)
HCD human-centred design
MVP minimum viable product
PDIA problem-driven iterative adaptation
PEA political economy analysis
TWP thinking and working politically
USAID US Agency for International Development

6
Executive summary
In uncertain and volatile contexts, where the and iterative prototyping to test options with
path to a desired outcome is not known up- users before committing in full. Lean startup
front, a linear plan-and-execute approach is provides a robust framing for learning and
unlikely to work. This is especially the case experimenting quickly where we know least.
with transformative change in complex systems. Teams test assumptions proactively with
However, when faced with complexity and prototypes, to identify whether users value a
disruption, it is tempting to hold onto what we product or service, whether there is a route to
know, reducing problems to fit the tools we have. scale and (when applied to social problems in
Adaptive approaches help by providing rhythms the form of lean impact) whether a product or
and processes for listening, learning, reflecting, service is helping people. Taken a step further,
making decisions and acting. These approaches these experiments could be embedded in a
have arisen in several sectors, for example in portfolio of initiatives which could together
software development (agile), product and service create the conditions for systemic change. TWP
design (human-centred design (HCD)) and remains important throughout, to understand
technology startups (lean startup, lean impact). the changing context and make politically-
In international development they take the form informed decisions.
of adaptive management, thinking and working For those working in international
politically (TWP) and problem-driven iterative development, having a diversity of approaches
adaptation (PDIA). to draw on, including those from outside the
Adaptive approaches help practitioners development sphere, is a good thing. Indeed,
grapple with uncertainty and complexity. development practice has begun to learn from
Using adaptive approaches can help counteract these methods, to open up more and earlier and
misplaced certainty in how problems are build in more diverse user feedback. The case
thought about and tackled. By talking to of PDIA, which is itself a fusion of some of the
potential users, understanding institutions, other approaches, shows that these approaches
interests and ideas and investigating the root can be fruitfully combined. However, while all
causes of a problem, practitioners using these of these approaches are valuable when used in
approaches help to illuminate the underlying the right context, practitioners may be perplexed
nature of the problem and the context. These by the multiplicity of methods and jargon. This
approaches help us to understand the system paper aims to address some of this confusion
in which we are working, and to continue to by mapping where these approaches have come
reflect on that throughout an intervention. from and showing how they can be applied across
Rather than building a whole solution the adaptive programme cycle. Armed with this
straight away, these approaches encourage us knowledge, practitioners might experiment with
to start small in order to learn and adapt more different combinations and sequences of adaptive
quickly (and cheaply). They commonly use approaches according to the kind of problem
structured cycles of testing and learning, with and context faced. In turn, this may help us
opportunities to gather and reflect together. move beyond a siloed view of approaches linked
HCD elevates human experience and creativity, to innovation, adaptive management or more
encouraging a wide set of potential solutions politically smart ways of working.

7
1  Introduction
The basic principles of the test-and- applied at the project or product level, they are
learn approach apply in almost any used by software developers, startup founders,
situation where people are trying to designers, civil servants, programme managers
solve problems in dynamic, uncertain and development entrepreneurs.
conditions (Berger, 2014: 122). Comparing adaptive approaches can be like
comparing apples and oranges. Some refer to
particular methods and tools, while others
This paper is about how to learn and adapt in embody more conceptual approaches that can
conditions of uncertainty. The approaches that be interpreted in many ways. In the development
it describes originate in different sectors but sector, approaches often overlap. Some have
have a number of shared principles. They are all fed off each other, sometimes indirectly and
used when one cannot determine the ‘correct’ sometimes explicitly, such as lean impact growing
course of action ahead of time. They all reject out of lean startup. This paper does not attempt
linear planning and execution, whether that is to provide a definitive classification. Instead,
‘waterfall’ project management in the software it aims to help make sense of some of the
world or blueprints copied from one country similarities and differences across and between
to another in the development sector without these approaches.
considering whether they meet local needs. This paper partly grew from internal
They all work in cycles of testing, learning and reflections within the LearnAdapt2 programme
adaptation, and often aim to engage with users team on bridging our own siloes and clarifying
early on. They are all responses to uncertainty what we can offer together. It is primarily aimed
and complexity. Lessons from one approach at those in the international development sector
might be applied in another context, and who are already practising adaptive management
approaches may be combined. and want to widen the possible approaches they
This paper compares six of the most draw on or make more informed choices in the
prominent adaptive approaches to emerge over future about which approaches to use, when and
the past two decades.1 Three come from the why. While all of these approaches are valuable
world of innovation, largely in the private sector when used in the right context, practitioners
(agile, lean startup and HCD), and three from may be perplexed by the multiplicity of
the global development sector (TWP, forms methods and jargon associated with adaptive
of adaptive management and PDIA). Mostly approaches.We are at risk of moving from the

1 There are a number of adaptive approaches this paper does not discuss in detail, including methods to explore future
scenarios, uncertainties and risks (foresight), systems practice and behavioural design (combining human-design methods
with experimentation and building on behavioural insights). These approaches overlap to a degree with the ones
discussed in detail. Behavioural design, for example, applies the experiences of behavioural insight and HCD practitioners
in designing context-fit and relevant experiments (Tantia, 2017).

2 LearnAdapt is a collaboration between the UK Foreign, Commonwealth and Development Office (FCDO), ODI, Brink
and Feedback Labs to explore how to manage adaptive development programmes better. It draws on approaches from the
development and tech sector including adaptive management, agile ways of working and lean startup. See www.odi.org/
projects/2933-learnadapt-innovation-and-adaptation-dfid.

8
methodological monoculture of the log frame to in conditions of uncertainty and volatility (Ries,
a paralysing Tower of Babel.3 This paper aims 2011a). On the development side, a significant
to address some of this confusion by showing body of evidence suggests that blueprint planning
how these approaches can be applied across the approaches can lead to large-scale failures in
adaptive programme cycle, from understanding conditions of uncertainty and complexity; instead,
problems to creating the conditions for scale and sustained change emerges from identifying locally
sustainable change. determined, context-specific and politically smart
As shown in the next section, the origins of solutions (Therkildsen, 1988; Ramalingam, 2013;
adaptive approaches tend to involve a shared Booth, 2015).
recognition of failure to grapple with complexity Pioneers developed adaptive approaches in
and uncertainty. For example, in software there different sectors to tackle uncertainty, rather than
was a long lag, approximately three years, ignoring it. Development practitioners can learn
between stating a project’s requirements and an from this wealth of experience and integrate
actual application being shipped. Over those some of it into their thinking and practice, be it
three years those requirements were likely to shift, throughout a project or in a more discrete way.
leading to software being obsolete on arrival The paper begins with an introduction to the
(Varhol, 2015). Similarly, conventional approaches origins of each approach, to set the scene. It
to business planning, market research and product then explores how to use practices from these
development do not work in the fast-paced approaches throughout the adaptive programme
world of technology startups. While spending cycle. This takes us from diagnosing a problem
months or years refining a product or service and understanding the context, through to
before sharing it with prospective customers implementing and learning via feedback
might work in an established industry with loops, to adapting with the end in mind. The
known customers, a startup taking this approach conclusion sets out how these approaches might
is likely to fail very quickly because it operates be fruitfully combined.

3 B. Ramalingam, personal communication, 2020.

9
2  Introducing each
approach
The concepts and terminology of innovation and so that the software was often out of date by the
adaptive approaches can be confusing, at least time it was released, and there were gaps between
at first. How is agile different to lean? How is what users needed and how these needs were
a prototype different from a minimum viable translated into software design (Varhol, 2015).
product? This chapter tells the origin story of each Alternative approaches began to emphasise
approach, explains its context and summarises incremental improvement (trial and error) and
core practices. The rest of the paper dives deeper creating working prototypes (Boehm, 1988). A
into working patterns, techniques and examples. major driver was the move from tangible product
to software delivered online. Developers suddenly
2.1  Agile got the ability to make continuous updates they
could deliver to customers in real time. In 2001, 17
Agile is a time boxed, iterative software experts gathered in Snowbird, Utah, and
approach to software delivery that created the Agile Manifesto, outlining 12 principles
builds software incrementally from the for building working software quickly and getting
start of the project, instead of trying rapid feedback from end users (Beck et al., 2001).
to deliver it all at once near the end This approach offered dual benefits: helping to
(Rasmusson, 2014). deliver software more quickly, and providing
the ability to test features and change course at
In the early 1990s, software development faced a an early stage. The Agile Manifesto combined
crisis. It was estimated that there was a three-year best engineering practices known as ‘Extreme
lag from start to finish in developing a product: Programming’ and an approach centred around
from stating a software project’s requirements to small teams working towards a common goal in
an actual application being produced and shipped iterative development cycles known as Scrum (a
in the form of floppy disks and CD-ROMs. Over reference to rugby). Agile teams ‘build quickly, test
those three years, these requirements – or indeed what they’ve built and iterate their work based on
the whole business – were likely to shift, especially regular feedback’ (UK Government, 2020).
as technology advanced. At the time, software was Today, surveys suggest that agile is now used
developed according to a heavyweight and linear by most IT professionals, whether this is ‘pure
methodology known as ‘waterfall’. This borrowed agile’ or a hybrid with traditional waterfall
from approaches to physical engineering. It is methods. This shift has happened in the last 10
called waterfall because teams complete one step years (Jeremiah, 2017). Agile is now applied to
before moving on to the next – flowing in one other types of projects and products far beyond
direction only. The prevailing wisdom among software. At the World Bank, a community of
software developers was that more time spent 200 ‘Agile Champions’ has been established as
planning at the outset would save money later on. part of an effort to create a culture of continuous
This resulted, however, in lengthy design processes improvement (World Bank, 2019).

10
2.2  Human-centred design and delivery. Divergent thinking means that
even unusual ideas are considered, and designers
Originating in product design, design thinking do not have a particular solution in mind from
involves generating and testing creative solutions the outset. Designers build tangible prototypes,
that people will adopt. Within this approach, such as models, videos or role-playing exercises,
HCD focuses on understanding the users of to generate conversation and get feedback
products or services and creating things which from potential users. This means even ‘failed’
are beneficial to them. Human perspectives prototypes are still useful as they facilitate
are considered at multiple points in the design convergent thinking in the second half of the
process, from observing what the problem is, design process. IDEO, a US design firm, uses
through coming up with ideas, to testing out a similar pattern of inspiration, ideation and
potential solutions. Establishing a personal implementation (IDEO Design Kit, n.d.). A
connection with users in order to see the world fourth ‘i’ might be iteration, as these stages are
through their eyes and gain a deep understanding run repeatedly to refine a prototype.
of their needs is therefore crucial. User-experience There are numerous examples of HCD being
design (often shortened as ‘UX design’) can be applied to social problems. In the UK, social
seen as a subset of HCD. This aims to improve entrepreneur Hilary Cottam has used design
the experience of the user, in terms of usability, thinking, along with participatory methods,
accessibility and pleasure, drawing on insights to reimagine the welfare state, launching
from psychology and other social sciences experiments in a range of sectors including health,
(Interaction Design Foundation, n.d.). ageing, family life and youth people (Cottam,
One way to visualise the design process is 2018). DFID, the US Agency for International
the UK Design Council double diamond (see Development (USAID) and the Australian
Figure 1). The first diamond is the process of Department for Foreign Affairs and Trade (DFAT)
problem discovery and definition, which results have applied HCD in programmes including
in a problem definition, or design brief. The SPRING (SPRING and fuseproject, 2019), an
second is the process of solution development accelerator programme for 75 businesses that

Figure 1  The UK Design Council double diamond

fin m
de oble

n
itio
Pr

Discover Define Develop Deliver

Behaviour-led Creative Review ideas Prototyping,
design research workshops and through culture selection and
idea generation thinking and design monitoring
Source: Design Council (2019)

11
served adolescent girls across nine countries in models while executing a known strategy, with
East Africa and South Asia. The programme, different processes for managing each.
which ran between 2014 and 2019, used a HCD Lean impact, developed by Ann Mei Chang,
process of research, storytelling and synthesis, former Google Executive and Chief Innovation
framing design challenges, brainstorming solutions Officer at USAID, is an approach to social
and prototyping those solutions. good based on the principles of lean startup.
The main principles are to think big (set goals
2.3  Lean startup based on the size of the need in the real world),
start small (to test and learn more quickly and
The goal of a startup is to figure out cheaply) and focus relentlessly on impact. Chang
the right thing to build – the thing acknowledges that social innovation is harder
customers want and will pay for – as than tech innovation given the social sector’s
quickly as possible (Ries, 2011a: 20). predilection for planning in advance and the
fact that those who pay and those who benefit
In earlier eras, business success would come have different interests. However, MVPs can be
from planning, market research and creating used to test hypotheses about impact. FCDO’s
and implementing the resulting strategy. This Frontier Technologies Hub programme uses
works well when forecasting is based on a stable a lean impact approach to help apply frontier
environment and a track record of operations, technologies to tackle development problems.
but does not work with startups because they The programme team has used short build-
operate with too much uncertainty around measure-learn loops to test hypotheses about the
who their customer is, or what their product potential for 3D printing in Nepal, tracking UK
should be (Blank, 2013). With the lean startup aid using Blockchain and introducing electric
approach, the job of a startup is to discover motorcycle taxis in Rwanda (Rahman, 2018).
the right products to build to provide value to
customers, and to develop a business model that 2.4  Thinking and working politically
works as quickly as possible in order to avoid
failure (Ries, 2011a). Rather than spending Capacity and technical knowledge
months or years refining a product or service alone are insufficient to change deeply
before sharing it with prospective customers, entrenched political interests and
startups should share early versions to get bureaucratic norms (Teskey, 2017: 2).
feedback and create a ‘build-measure-learn’
loop to test the riskiest assumptions behind the TWP is an approach to development
business model. Lean startup advocates for the interventions that entails thinking in a more
use of ‘minimum viable products’ (MVPs) – that politically aware way – for instance through
is, building a basic model of the new product to political economy analysis as an ongoing process
be tested with customers before any large-scale or mindset – and working differently as a result,
investment. For example, a company might share in ways that are tailored to contextual realities
a crowdfunding campaign, sign-up sheet or and that call for flexibility and adaptation
video explainer for a product they have not yet (Rocha Menocal, 2014). While there is no
created to test demand – this is how Dropbox single agreed definition or framework, the TWP
started (Ries, 2011b). If the MVP is successful, Community of Practice (2013) sets out three
it can be refined. If it fails with customers, the core principles: ‘Strong political analysis, insight
startup knows to change direction. and understanding; a detailed appreciation
More recently, this approach has been adapted of, and response to, the local context; and
for larger organisations. Ries (2017) outlines flexibility and adaptability in program design
principles of entrepreneurial management in The and implementation’. TWP starts with a
startup way, while The corporate startup (Viki et recognition that developmental change processes
al., 2017) argues for ‘ambidextrous’ organisations are inherently political, and that development
that are capable of searching for new business programmes are therefore more likely to be

12
successful when they consider, and have the strategic and tactical reviews allow space to
flexibility to adapt to, local political dynamics course correct and scale up what works, a
(Hogg and Leftwich, 2008). These include the process supported by continual and timely
formal and informal ‘rules of the game’, the evaluation, context monitoring and learning.
power and interests of different leaders and Adaptive management also emphasises the
groups, and ideas, norms and values. Working importance of locally-led problem-solving,
politically means that development actors not meaning that change is led by those within the
only tailor interventions to local conditions, but context rather than being externally driven
also consider themselves political agents in their (Wild et al., 2017). To understand the local
own right, and therefore part of the context context and underlying politics, in practice
(Laws and Marquette, 2018). adaptive management is often supported by
Working politically may include acting in a processes of TWP.
politically smart way by supporting or facilitating The principles of adaptive management were
coalitions and working alliances. Because shifting articulated by a group of practitioners in the
incentives is a complex undertaking that involves ‘Doing development differently’ manifesto
altering power relations, development actors (2014), and have gained currency across the
need strong processes for understanding, testing sector since then. Several aid agencies have
and learning, often with a focus on incremental launched initiatives to integrate adaptive
and small-scale reforms at first. The emphasis is management into the way they work, including
on continual analysis to understand the changing LearnAdapt at FCDO, USAID’s Collaborating,
political context and make politically informed Learning and Adapting framework, the Global
decisions, rather than producing a weighty Delivery Initiative at the World Bank and Global
upfront report. This is typified in ‘Everyday Learning for Adaptive Management, a joint
political analysis’, a bare bones framework to FCDO–USAID initiative.4
help frontline staff make politically informed
decisions (Hudson et al., 2016). Working 2.6  Problem-driven iterative
politically is often operationalised through adaptation
adaptive management.
PDIA is an approach to adaptive management,
2.5  Adaptive management most often used in government reform processes.
It shares DNA with almost all of the approaches
USAID (2018a) defines adaptive management discussed above, featuring the time-boxed iterations
as ‘an intentional approach to making decisions of agile, the ideation phase from design thinking
and adjustments in response to new information and the political understanding of TWP. It was
and changes in context’. Adaptive management first tried in Mozambique in 2009 (Andrews et al.,
therefore legitimises changes in tactics and 2018). The Building State Capability programme
strategy as part of a deliberate approach. While at Harvard has pioneered PDIA, with a number of
aid programmes using adaptive management projects around the world. The team has shared the
should have clear goals, the pathways to those approach in various formats, including courses on
goals are not easily defined ahead of time, the practice of PDIA, a book (Andrews et al., 2017)
meaning that activities and outputs are not and a toolkit (Samji et al., 2018).
specified upfront. Instead, programmes build PDIA is a response to stubbornly low levels
in deliberate processes of testing, learning and of capability of developing country governments
experimentation throughout delivery to discover and the failure of aid programmes that have
what will work most effectively. These regular attempted to reform these institutions. When aid

4 Learn more about these initiatives: LearnAdapt (www.odi.org/projects/2933-learnadapt-innovation-and-adaptation-

dfid); Collaborating, Learning and Adapting (www.usaidlearninglab.org/cla-toolkit); Global Delivery Initiative
(www.globaldeliveryinitiative.org) and Global Learning for Adaptive Management (www.odi.org/projects/2918-global-
learning-adaptive-management-initiative-glam).

13
programmes transplant ‘best practice’ formal 6. Be intentional about learning, using research
institutions from outside, reforms mimic the and prototypes to test hypotheses.
appearance of change but have little anchoring in 7. Measure primarily to learn, rather than to
contextual realities and lead to very little change report.
in functions. PDIA begins by defining the problem 8. Have regular junctures for reflection and
(rather than starting with an imposed solution), learning.
and deconstructing it to get at the root causes, 9. Work in loops instead of in a straight line, so
with those leading the reform at the core of the that planning, implementation and learning
process. Ideas for solutions are often identified are no longer separate processes.
within context, rather than importing external 10. Be pragmatic about process; do what’s
‘best practice’. Multiple potential solutions are needed, not what looks best or is considered
then tested in short cycles of action and reflection, ‘best practice’.
which are repeated until the problem is solved.5
There are also some differences, as would be
2.7  Common principles of adaptive expected for approaches that evolved in different
approaches contexts. In particular, development approaches
(TWP, adaptive management and PDIA) place
As described above and as shown in Annex 1, a greater emphasis on power and politics,
some principles are common across all these which is largely absent from more technocratic
approaches: approaches like agile and lean startup. Adaptive
approaches in development have a wider range
1. Acknowledge that the answer is not (and of options for what to create and facilitate – not
cannot be) known upfront, and that there may just products or services, but also forms of
not be a single answer. collective action. In addition, the idea of scaling
2. Recognise the importance of the political, can be different. For the private sector, scaling
social and economic context to understand tends to mean creating a product or service that
any given problem or issue. can be replicated and sold to the mass market.
3. Start with the people you’re building for or The development sector tends to start at the
working with, and encourage participation level of a problem and what it would take to
and listening. sustainably address it – which might come from
4. Recognise that understanding a complex large-scale roll-out, but could equally come from
system or problem requires interacting with it. finding a leverage point in the system or creating
5. Start small, with ‘little bets’ that incur low the conditions for growth, as explored at the end
costs for failure. of the next chapter.

5 For more see the Building State Capability programme at the Harvard Center for International Development
(https://bsc.cid.harvard.edu).

14
3  Adaptive approaches
throughout the
programme cycle
People working on adaptive development but also to ways of working and how to structure
programmes will be familiar with a programme an organisation. Agile organisations typically
cycle where they assess and design, implement involve small, cross-functional teams that learn
and adapt, making tactical and strategic changes and evolve together, as at organisations such as
along the way. In most scenarios, using a single Spotify, Skyscanner and MercyCorps (MercyCorps,
adaptive approach from end to end is not optimal; 2017; Freeman, 2018). Agile processes are used in
there is no new best practice for all seasons. the UK by the Government Digital Service (GDS)
Instead, practitioners must engage in bricolage, (UK Government, 2019). GDS applies different
combining and modifying different approaches governance principles (e.g. ‘don’t slow down
for their needs. While this places a higher burden delivery’, ‘decisions where they’re needed at the
on practitioners to understand and apply a variety right level’) and uses different, more streamlined,
of methods and ideas, it makes it more likely that procurement and contracting approaches to the
there will be a better match between the problem earlier, ‘learning’ phases of agile projects (HM
at hand and the way it is addressed. What Treasury, 2018).
follows is a guide for those unsure about which The political economy of aid is not the focus
approaches to apply when faced with conditions of this paper. However, it remains the case that
of complexity and uncertainty. Hooking elements international development actors often find it
of adaptive approaches to the familiar programme very difficult to work in adaptive ways because
cycle may encourage practitioners to experiment the environment in which they operate does not
and expand their toolbox and ways of working provide the space and incentives to do so (see
(see Figure 2). for example Natsios, 2010; Rocha Menocal,
2014b; Valters and Whitty, 2017). For example,
3.1  Enabling adaptation organisations implementing aid projects face
pressure to disburse funding to achieve pre-
Experience from the private and development determined results according to a preset timeline.
sectors tells us that culture, mindsets and skills Accountability tends to be upwards to donors.
can often be more important than the adaptive The constraints of the aid world might mean that
approach chosen.6 Practitioners share a recognition embracing a particular adaptive approach or
of the importance of the wider enabling combination of approaches would not be feasible.
environment; these approaches do not work as ODI and FCDO are currently undertaking
well in an environment of more mechanistic, research into adaptive bureaucracies to explore
linear thinking. For this reason, the term ‘agile’ what characterises organisational environments in
refers not only to working with software itself, which adaptation is possible and encouraged.

6 See Goeldner Byrne et al. (2016) and Haugh and Salib (2017) for reviews of this covering adaptive management, and
Edmondson (2010) on teams and culture in conditions of uncertainty more broadly.

15
Figure 2  Adaptive approaches through the programme cycle

Diagnosing the Understanding

problem context: power, politics
• Discovery Phase, e.g. and systems
immersion (HCD)
• Political economy
Adapt with the end in • Problem construction,
analysis, institutions,
mind problem deconstruction
ideas and interests
(PDIA)
• Think big, start small; (TWP)
test growth hypothesis • Change space analysis
(PDIA) Cycles of testing and
(lean startup)
• Innovation portfolio (lean learning
• Look for leverage and
Enable alternative endgames startup)
• Systems prototyping
• Sprints, scrum
• Culture,
mindsets,
• Contention and
collective action Assess and teams, daily standup,
retrospectives (Agile)
skills, structure
• Adaptive
• Create adaptive capacity
design • Push periods (PDIA)

bureaucracies
• Agile
organisations Adapt Implement
Continual reflection Learn through
and action on power Measure to learn and Seek and act upon experiments
and politics adapt feedback • Prototypes (HCD)
• Innovation accounting • Design sprints, close the • Minimum viable products,
• Collective action, test value and impact
coalitions (TWP) (Lean startup) loop (HCD)
• Open-ended but • Constituent feedback hypotheses (Lean
• Build authorising space startup/impact)
(PDIA) countable metrics, (AM)*
developmental • Portfolio of multiple
• Behavioural design experiments, strategy
evaluation (AM)*
testing (AM)*

Source: Based on Ramalingam et al. (2019: 9)

Note: AM is adaptive management

3.2  Assess and design articulate and may only emerge once initial work
has been delivered.
3.2.1  Diagnosing the problem An important way adaptive approaches can
In some contexts, there may be agreement on help development practitioners is by challenging
what the problem is, grounded in evidence, over-confidence in a single interpretation
but it is unclear what the potential solutions of what the problem might be, and reliance
might be. In these circumstances, there is a need on pre-identified solutions or blueprints. In
to test, learn and collaborate. At other times, government reform, where the problem is loosely
the nature of the problem itself is unclear or defined, the donor determines a solution, and the
contentious, suggesting a need both to determine government acquiesces by building something
the underlying nature of the problem, and to that looks like the donor’s solution, but which
develop ways to address it at the same time. functions more like the status quo (known as
Agile delivery is designed to tackle the first type ‘isomorphic mimicry’) (Andrews et al., 2017).
of problem. While there is some uncertainty Methods from HCD, TWP and PDIA can help
about what the product will end up looking like, practitioners acknowledge complexity, counteract
the challenge at hand is relatively structured. misplaced certainty in a problem definition and
The client gives the software development team avoid importing blueprint solutions that are not
a set of requirements, and the team responds grounded in contextual realities.
to this, delivering the highest-value work first. HCD can help illuminate the underlying
There will, however, be a need for flexibility nature of the problem, while developing
because what the client needs may be difficult to solutions. In the discovery phase, the problem is

16
explored and defined by getting a range of inputs context. For instance, HCD can include a period
from interviews, observation and other sources, of immersion in a community. However, Conway
and synthesising these as a team. This somewhat et al. (2018) suggest that the conventional double
sidesteps the issue of disagreement in the diamond model of design thinking (defining
community the team is designing for; the design the problem then developing the solution)
team is ultimately responsible for synthesising is not equipped to deal with complex social
information and deciding how to define the challenges. They suggest ‘think[ing] like a system,
problem. Where there is mild disagreement this act[ing] like an entrepreneur’ (Figure 3). By
might work well, though it is subject to the starting with framing and mapping the system,
biases, blind-spots and privileges of the design practitioners can better understand the problem
team. With more sensitive problems, this process and its context, before spotting opportunities,
is likely to require a more participatory approach prototyping and navigating barriers in an
that helps a group come to their own conclusions entrepreneurial way. This is akin to the early
through dialogue. stages of PDIA, in which the context is explored
Where there is great disagreement between through the lens of ‘constructing’ the problem.
groups on both the problem and potential With a systems lens, it is possible to identify
solutions, we enter the realm of politics. the most effective entry point into a problem
Aid programmes aimed at reform may fail – perhaps a part of the system that is ripe
because powerful groups benefit from existing for action or influencing (Meadows, 1999).
arrangements and resist change (Leftwich, 2011). This allows us to be more intentional about
TWP and PDIA include methods to understand designing a portfolio of experiments and
and engage with problems that are deeply potential interventions. This has been called
political in nature. Political economy analysis ‘systems prototyping’: ‘experimenting with a
prior to an intervention is now a standard part range of opportunities for change, but with a
of the toolbox of most aid programmes (see, for coherent (and plural) narrative that links them
example, USAID, 2018b). TWP means that this is all together’ (Robinson, 2019). An innovation
not a one-off exercise but a continual process of portfolio can include more disruptive elements
understanding the political, economic and social alongside incremental improvements.
context. In PDIA, constructing a problem as a TWP is likely to be especially helpful in
cross-government team means showing the need understanding the institutions, power, interests
for real change, creating political momentum for and ideas behind stubborn development
reform. The problem is then deconstructed to get challenges by prompting practitioners to ask
at root causes. Once these are identified, teams questions about beliefs, norms, constraints,
look at the ‘change space’ for each one. Is there power dynamics and the space for change.
support from those who can authorise change? Is Importantly, this is not a one-off exercise
the need for change accepted by those who will – more information becomes available as
be affected by it? Does the team have the abilities an intervention progresses and the context
needed to undertake an intervention? Once a continues to change. Perfect certainty is never
problem is better understood, agile could be used possible, so interventions need flexibility and
for delivering an intervention. real-time feedback. The humility at the heart of
adaptive management, and the opportunities it
3.2.2  Understanding context: politics, gives to pause, reflect and change course, means
power and systems that continual monitoring of the environment
Without properly considering context, can feed into decision-making. In stable
interventions and innovations are likely to hit situations, uncertainty about context can be
barriers to change, such as competing incentives, reduced through good exploration and analysis.
cultural norms, power dynamics and the For development challenges in more volatile
informal institutions that shape how regulatory and opaque contexts, uncertainty can only be
frameworks operate. The approaches featured reduced (though not eliminated) for a moment in
here offer a number of tools to better understand time, but not permanently. For instance, political

17
Figure 3  Think like a system, act like an entrepreneur

Think like a system Act like an entrepreneur

Divergent Convergent Divergent Convergent

thinking thinking thinking thinking

`
Problem
definition

Discover Define Develop Deliver

Behaviour-led Workshops Review ideas Prototyping,
design research and ideas through culture selection and
generation thinking and mentoring
design
Source: Conway et al., 2018

economy analysis can help in understanding a feedback from users and act upon it, what to
given setting in its full complexity, but cannot measure to inform decision-making and how to
eliminate uncertainty. It can reveal constraints, continually reflect on power and politics.
but does not identify a clear way forward.
This explains donors’ disappointment with 3.3.1  Working in cycles of testing
political economy analysis (for example and learning
‘Drivers of change’ thinking in FCDO). Donors Adaptive approaches all reject linear planning
were looking for a tool that would remove and execution as the main form of delivery.7 In
uncertainty altogether and enable them to their place, they promote working incrementally
programme in waterfall mode – they were on smaller chunks through iterative cycles of
perhaps thinking like a system, but not then experimentation, testing and reflection – working
acting like an entrepreneur. in loops instead of a straight line. This helps to
verify that a team’s thinking is sound and uncover
3.3  Implement and adapt operational challenges early on, based on the
principle that a team can only really understand a
By their nature, adaptive approaches do not complex system by interacting with it.
separate implementation and adaptation into Agile offers more structured processes
neat components. However, we can derive some for testing and learning, working in shorter
helpful principles. We can learn how to create a production cycles known as sprints. Sprints are
rhythm for cycles of testing and learning, what time-boxed iterations during which specific work
to do first to learn most quickly, how to seek out has to be completed – these might last about two

7 Some elements of adaptive projects are likely to be more predictable and therefore well-suited to a more linear process.
The important thing is to diagnose this and respond to each element accordingly.

18
weeks. Teams aim to deliver high-value work first. hypotheses about what customers are willing
Teams are self-organising and cross-functional, to pay for (the value hypothesis) and whether
with three main roles in the dominant Scrum there is a viable route to growth (the growth
approach: Product Owner (who liaises with the hypothesis). For example, a company might
client, defines the vision for the product and launch a crowdfunding campaign, sign-up sheet
prioritises accordingly); Scrum Master (similar or video explainer for a product they have not
to a project manager who oversees operations yet created, in order to test demand. Importantly,
and the process); and team members (who create the assumptions tested first should be the riskiest
the products). There are regular face-to-face ones – those that would break the business model
meetings to keep everyone in sync, including a if they turned out to be false. ‘The goal is to
‘daily standup’ for progress updates and facilitated eliminate the greatest degree of risk with the least
‘retrospectives’ at the end of a sprint, where the investment of time and money’ (Chang, 2018).
team reflect on what went well and what could Assumptions that have been tested successfully
be improved, and plan for the next sprint. These are said to be ‘validated’.
rhythms create space to intentionally pause and Chang (2018) advocates a similar approach
reflect; without them, a team may default to for testing an impact hypothesis in the social
rushing ahead with whatever they are working sector. Instead of assumptions in a business
on, even if this is ineffective, or work towards model, this can relate to steps in a theory of
milestones which are no longer relevant. PDIA action – describing how an intervention is
uses ‘push periods’, based on agile sprints. designed to lead to a desired change. Are we sure
These include regular planned junctures for one thing will lead to another, and what evidence
reflection and learning, meaning that planning, do we have for that? If evidence is limited,
implementation and learning are no longer experiments could be designed to test these
separate processes. These cycles can be as short causal relationships, particularly between inputs,
as a couple of days – representing the smallest outputs and outcomes. This approach could be
portion of work necessary to generate useful used in conjunction with strategy testing, an
learning – but will tend to become longer as work approach developed by the Asia Foundation
progresses. The length of sprints or push periods to meet the learning and accountability needs
can also be dictated by how quick and easy it is to of adaptive programmes working on complex
make and deliver regular ‘updates’ or iterations. problems (Ladner, 2015). In this quarterly
structured discussion, the programme team look
3.3.2  Learning through experiments back at what has happened since the last meeting
Another principle that adaptive approaches have and review the theory of change in light of what
in common is to start small, with ‘little bets’ that they have learnt about the problem, key actors
have low costs for failure (Sims, 2011).8 Having and political, social and economic dynamics.9
a portfolio of multiple experiments means that This provides an opportunity to consider the
teams are less attached to particular solutions, health of the whole system, rather than of a
making it easier to change direction or stop. If a single intervention. If the underlying assumptions
lot of resources are put into one experiment, it are no longer valid, the team adjust the theory
becomes difficult to shut it down. In HCD, this is of change, and the programme strategies change
achieved by making prototypes which are tested accordingly. Setting out to test some assumptions
with potential users. explicitly through prototypes or MVPs could
For lean startup, learning is highly intentional. make this process faster and more intentional,
Startups create MVPs, built solely to test helping a team to make ‘little bets’ to test out

8 Working in smaller chunks does not mean that ambition is small. Chang (2018) combines the principles of ‘think big’ and
‘start small’ in the Lean Impact approach, also phrased as ‘nail it before you scale it’.

9 ‘A Theory of Change is an ongoing process of reflection to explore change and how it happens – and what that means for
the part we play in a particular context, sector and/or group of people’ (James, 2011: 3).

19
potential tipping points in a system, as well as development sector. Much development practice
failure points.10 If this technique was applied remains quite closed and could learn from
to advocacy, for instance, a ‘minimum viable innovation methods to open up more, earlier,
campaign’ could be used to test assumptions build in much more diverse user feedback and
about the best ways to achieve wider goals. use this information to make decisions (Jean,
Agile, lean startup and HCD have evolved 2017). While there is a history of constituent
mainly to create products and services; feedback in development, in practice this has
practitioners do not typically aim to challenge often been tokenistic or about demonstrating
power dynamics. Adaptive approaches in results rather than leading to significant course
development have a wider range of options for correction based on this feedback (Anderson
what to create: not just products or services, et al., 2012); it has not been very clearly linked
but forms of collective action such as coalitions, to adaptive management processes.11
movements or campaigns. Approaches such as Incorporating and valuing feedback requires
PDIA and TWP involve bringing together those cultural change (Anders, 2016). HCD offers a
with a stake in solving a complex problem and structured way, not only to hear from ‘users’ but
helping to facilitate a process to collectively also to co-create better solutions and ‘close the
create and test potential solutions. They may loop’ by showing how feedback has led to change.
‘work with the grain’ of political institutions and However, there are additional power imbalances
interests in order to make progress where there to contend with when applying these approaches
is support, but project teams tend to be highly to development challenges, with longstanding
conscious of these dynamics. The size of reform biases in the sector dividing stakeholders into
coalitions needs to be fit for purpose and, in an decision-makers and constituents and valuing the
echo of lean thinking, ‘it is usually best to have perspectives of experts above others (Chambers,
the smallest size necessary to achieve the goal’ 1997). For some development professionals, being
(Leftwich, 2012: 22). adaptive is a means to the end of shifting power
away from aid donors and towards the people
3.3.3  Seek out and act upon feedback the money is intended to help, both because this
Adaptive approaches emphasise ‘getting out makes programmes more effective and because
there’ to meet customers, clients and constituents this is valuable in its own right. Three of the six
to find ideas and get feedback. For instance, principles in the ‘Doing development differently’
Google Ventures uses ‘design sprints’ – a five-day manifesto centre on local leadership. However,
process to answer business questions through some have criticised this as the weakest element in
mapping problems, quickly sketching solutions, existing development reform efforts (Muyumbu,
developing a prototype and testing it with 2018). Indeed, it has been easier for aid agencies
users to see if live reactions validate hypotheses to adopt more technical aspects of adaptive
(Knapp et al., 2016). This is distinct from management than to make more fundamental
conducting more expensive randomised control power shifts (Rocha Menocal, 2014a).
trials, statistical surveys or academic research,
which can produce robust evidence that arrives 3.3.4  Measure to learn and adapt
too late to act upon. Rapid feedback loops in Adaptive approaches measure primarily to learn
agile and lean startup are often made possible by rather than to report, which is a significant
new technology – releasing in the cloud rather contrast to conventional models of measurement
than sending out software in boxes. Traditionally, in development practice. Adaptive approaches
this kind of feedback is not available in the have a shared critique of these traditional

10 As a word of caution, the experience of the LearnAdapt programme suggests that it is not always possible or easy to
answer a question with a prototype. It is often necessary to break down assumptions into smaller, short-term hypotheses
(see Sparkman, 2015).

11 A forthcoming LearnAdapt paper explores how constituent engagement and adaptive management are combined in practice.

20
measures: traditional business metrics such as based on implementation experience (Booth,
internal rate of return or net present value are 2018). These measures are helpful for telling
seen to make little sense for a completely new the story of what the programme team learned
product or service; similarly, ‘vanity metrics’ – the and what they are doing differently as a result.
number of people attending an event, the number Developmental evaluation is another approach
of hits on a website or the amount of money that can support innovation and adaptive
spent – might make a project look good, but do management in complex environments (Patton,
not help us understand whether it is going in the 2010). Measures and tracking mechanisms are
right direction or changing anything in substance. developed quickly as outcomes emerge, and can
There is often a similar tendency to count and change over time. The emphasis is on supporting
report unhelpful measures in development decision-making in real time. This makes it well-
projects in the name of accountability to a suited to situations where both the solution and
funder. Much of the time, these measure activity the context are changing quickly. Once a solution
rather than progress in solving a problem, is better-established, it may be riper for robust
creating perverse incentives (Natsios, 2010; assessment using a randomised control trial or
Valters and Whitty, 2017). similar method.
Innovation accounting, as developed for
lean startups, uses different metrics which 3.3.5  Continual reflection and action on
are more actionable. This might include the power and politics
number of new ideas generated, the percentage Burns and Worsley (2015) describe
of these which have been tested with MVPs and transformative change as ‘essentially change
validated, and the speed of learning (Viki et al., which shifts power relations’. Any intervention
2017). Metrics may also be more useful at the aimed at transformative change is more likely
unit level rather than in aggregate; the percentage to rub up against the incentives and goals
of customers who upgrade from a free product of powerful actors and institutions, creating
to a paid version or the percentage of users who barriers to change. Even an innovative and
recommend a service to others is more useful technically brilliant service or product could be
information at an early stage of developing a blocked or made irrelevant by wider systemic
business model than the total number of users. and political constraints. While private sector
Unit-level metrics can demonstrate that users innovators must adapt to changing ideas and
value the product or service, and that there is preferences, the shifting whims of clients and
a path to growth and sustainability, therefore markets are of a different nature to a turbulent
validating the value and growth hypotheses. political environment.12 Adaptive development
An adapted version of innovation accounting programmes are not just a process of discovery;
could support adaptive approaches in they have to consider that what works is often
development. Some adaptive programmes are hindered by underlying incentives, institutions
starting to use equivalent metrics. For example, and power dynamics. An effective development
FCDO’s Economic Policy Incubator programme project is not just about chasing the changing
in Nepal uses open-ended (i.e. not dictating needs and desires of customers; it’s often more
outputs) but countable metrics as part of its akin to actually influencing multiple competing
PDIA approach. Outcome indicators relate to interests, or at least trying to find a way to
steps towards solving the problem at hand, such work through the mess of all these conflicting,
as ‘Number of top 9 cross-sectoral constraints changing, imperceptible interests.
to inclusive and transformational growth This is where it is especially important
significantly eased (cumulative)’, while output to think and work politically. For instance,
indicators include agreeing project plans with ‘everyday political analysis’ helps people
partners and the percentage of initiatives adapted understand interests and change in their daily

12 While agile, lean startup and HCD in themselves do not address political turbulence, companies constantly monitor this
in other ways and actively participate in the political arena in the form of lobbying, donations and public relations.

21
work (Hudson et al., 2016). In PDIA, teams Instead, we might consider how to create the
take a proactive approach to gain the support conditions for growth. This could take a number
of authorisers such as senior civil servants of paths:
or ministers who might otherwise block an
initiative or be indifferent. By communicating • Look for leverage, and then look again: To
early wins, reformers gain legitimacy and have create transformation, we should search for
further authorising space to experiment and opportunities where a relatively small effort
refine their solutions. A behavioural design could lead to large change and open up new
approach, applying insights from behavioural possibilities. TWP and PDIA provide tools
science rather than relying on intuition, could to undertake this diagnosis. These should
supplement political thinking (Tantia, 2017). be used continually. Working to scale up
a specific solution could blind innovators
3.4  Adapt with the end in mind to other, more catalytic roles they could
potentially play. Thus, strategy testing can be
Adaptive approaches in the private sector and in used to examine assumptions behind change
development can show us different ways to think efforts on a regular basis as the system itself
about scale and systems change. evolves (Ladner, 2015).
Rather than seeing growth on a linear path, • Consider how others might take up an
lean impact challenges us to think big but idea: An intervention can grow in scale and
start small (Chang, 2018). Thinking big means impact without an organisation needing
considering what it would take to genuinely to grow with it. Gugelev and Stern (2015)
address the problem, rather than just a small list six ‘endgames’ or roles to play in the
part of it. Having a clear ambition means that overall solution to a social problem: open
the pragmatism of adaptive approaches can source, replication, government adoption,
be focused on the end goal, rather than on commercial adoption, mission achievement
perfect process. However, the flipside to this is and sustained service. Only the final option
that staying small for longer makes it easier to implies continued and increased operations
learn more quickly. Furthermore, retaining an as a project or organisation. Ideas from lean
experimental approach with multiple solutions startup can help to test for potential demand
at once can help a team stay humble about any from other stakeholders. For instance,
particular answer to a problem. Lean startup FCDO’s Frontier Tech Livestreaming
involves testing whether a product or service programme uses ‘yes, if’ sprints. The
could be taken to scale at an early stage. As programme team asked potential investors
well as using MVPs to test whether customers in electric motos in Rwanda what exactly it
would pay for a product or service, lean startups would take to invest in the idea (Rahman,
also test the ‘growth hypothesis’ of their route 2019). This helped with decision-making
to reaching more customers. For example, will about how to use resources. Depending on
users recommend the product to others? What how much contexts vary, interventions might
will the mass market pay for? Is it possible to be copied exactly or altered substantially to
reduce costs? make them more culturally appropriate. HCD
Systems thinking tells us that, in a non-linear could be used in this process of adaptation
world, the size or scale of an intervention is to quickly understand the needs of potential
not necessarily a useful guide to the scale of its users, build prototypes to test in the new
impact. In a complex system, ‘scale’ may be the context and create rapid feedback loops.
wrong metaphor to use as it suggests replicating • Contention and collective action: Complex
or increasing the size of a particular intervention. problems are very rarely solved by a

22
 single intervention or by one actor.13 It to deal with complex problems. Indeed,
takes a combination of smaller changes Ang (2017: 4) suggests that the adaptive
and people working in coalition to create development community must go beyond
sustainable change. Yanguas (2018) suggests ‘dispensing obvious advice like “avoid
we think of change processes in terms of mimicry” and “promote innovation”.
contention, rather than diffusion of an idea: The relevant question is instead, “How
‘Development is not a single process of can we create conditions that enable
change, but a tug-of-war between reform adaptation?”’. The best way to build this
mobilisation and demobilisation’. This capacity is to successfully move through an
implies that a linear path to scale is less adaptive challenge (Heifetz et al., 2009).
relevant in more political contexts, and For aid programmes, this requires working
collective action becomes more important. with people within the system, and is a
Coalition theory tells us that the optimal size challenge akin to a company helping other
of coalitions depends on the sort of change organisations to undertake design thinking
they are trying to bring about (Leftwich, for themselves.
2012). Wide-ranging social change probably
does require a broad coalition, but a coalition Finally, perhaps we should not always seek
to drive a specific policy change can be to scale. Small can be beautiful. Development
narrower, more temporary and consist of a actors often work on difficult problems in
critical mass of well-connected individuals. difficult places where even a rigorous process of
Having different possible endgames beyond searching will not result in a single solution that
scaling an intervention helps in thinking can be scaled up and replicated. In particular, a
through what different forms of success commitment to ‘leave no one behind’ requires
might look like. deep interventions that may vary significantly
• Create adaptive capacity: Challenges from place to place.14 Another form of success
such as providing high-quality education that does not involve scaling is ‘mission
or healthcare across a country require achievement’, where an intervention is redundant
governments, development actors and the because the original problem it was addressing
private sector to have the adaptive capacity has been solved.

13 Equally, this means that a donor should not try to address every aspect of a problem through different projects. Tackling
complex problems does not necessarily entail a highly complicated intervention with many parts. See Sharp et al. (2019)
for a full discussion on this in the context of DFID.

14 Positive deviance is one adaptive approach that can work well in such contexts. See Pascale et al. (2010).

23
4  Conclusion
The approaches discussed in this paper have More commonly, though, where development
evolved in different contexts, but have a great deal challenges are about systemic change and
in common. The historical usage of a method does changing behaviours and incentives (like any
not in and of itself determine the scope of contexts public policy challenge) in a political arena,
in which it can be valuable. For those working private sector approaches need to be adapted.
in international development, having a diversity In these cases, they may need to be used in
of approaches to draw on, including those from conjunction with elements of TWP, systems
outside the development sphere, can be very useful. thinking and adaptive management. PDIA
Given the increased recognition of and focus on is one such adaptation – applying agile and
complex problems in uncertain environments, lean principles to system change and technical
adaptive approaches are particularly important. By assistance in government systems.
looking at the roots of such approaches, comparing Adaptive aid programmes should experiment
them and mapping them onto the adaptive cycle, with different combinations and sequences of
this paper has sought to show how international adaptive approaches according to the kind of
development actors can draw from them, and make problem and context faced. PDIA is an off-
practical use of them. the-shelf example of this, but it is not the only
These approaches can be frameworks to possible combination. Lean impact adds a social
help practitioners to think in terms of systems impact element to the lean startup approach,
(including political dynamics) and act like though it does not seem well-suited to more
entrepreneurs. They nudge us to take context politically turbulent environments. What might it
seriously, test our ideas in the real world, look like to try TWP and lean impact together?
listen more intently and be humbler. Private How about HCD and adaptive management?
sector adaptive approaches such as agile, lean Could MVPs be used in conjunction with
startup and HCD can be especially helpful strategy testing to interrogate assumptions
for experimenting more quickly to find a more explicitly? The prize is to try this at
creative solution to a development challenge – intervention level. There is no inherent reason
particularly if this is a product or service. This is why elements of agile, lean startup and HCD
especially relevant for development programmes could not be used in systemic change efforts, if
that work through private sector or market they are embedded as single experiments in a
approaches. For these, private sector approaches portfolio of initiatives that is designed to shift
work with little alteration, and can often be the system. These experiments can be products or
fruitfully combined; the different methods serve services but can also be coalitions. Documenting
different purposes. For example, practitioners such combinations so as to better understand
use HCD to surface and understand problems what works in which contexts could provide
and user needs, and lean startup to build and international development actors like FCDO
test solutions to those problems. Lean startup with an invaluable opportunity for learning and
can combine with agile for testing and building. improving development practice.

24
References

Anders, M. (2016) ‘DfID’s evolving approach to beneficiary feedback’. Devex, 3 November

(www.devex.com/news/dfid-s-evolving-approach-to-beneficiary-feedback-89059).
Anderson, M., Brown, D. and Jean, I. (2012) Time to listen: hearing people on the receiving end of
international aid. Cambridge MA: CDA Collaborative Learning Projects.
Andrews, M., Pritchett, L. and Woolcock, M. (2012) Escaping capability traps through problem
driven iIterative adaptation (PDIA). CID Working Paper No. 240. Cambridge MA: Center for
International Development.
Andrews, M., Pritchett, L. and Woolcock, M. (2017) Building state capability: evidence, analysis,
action. Oxford: Oxford University Press.
Andrews, M., McNaught, T. and Samji, S. (2018) Opening adaptation windows onto public financial
management reform gaps in Mozambique. CID Faculty Working Paper No. 341. Cambridge MA:
Center for International Development.
Ang, Y.Y. (2016) How China escaped the poverty trap. Ithaca, NY: Cornell University Press.
Beck, K. , Beedle, M., van Bennekum, A., et al. (2001) ‘Principles behind the Agile Manifesto’. Webpage
(https://agilemanifesto.org/principles.html).
Berger, W. (2014) A more beautiful question: the power of inquiry to spark breakthrough ideas. New
York: Bloomsbury.
Blank, S. (2013) ‘Why the lean startup changes everything’. Harvard Business Review, May
(https://hbr.org/2013/05/why-the-lean-start-up-changes-everything).
Boehm, B.W. (1988) ‘A spiral model of software development and enhancement’ Computer 21(5): 61–72.
Booth, D. (2015) ‘Still watering white elephants? The blueprint versus process debate thirty years on’
in A.M. Kjær, et al. (eds) Perspective on politics, production, and public administration in Africa:
essays in honour of Ole Therkildsen. Copenhagen: Danish Institute for International Studies.
Booth, D. (2018) Incubating policy for economic transformation: lessons from Nepal. Report. London: ODI
(www.odi.org/publications/11090-incubating-policy-economic-transformation-lessons-nepal).
Burns, D. and Worsley, S. (2015) Navigating complexity in international development. Rugby:
Practical Action Publishing.
Chambers, R. (1997) Whose reality counts? Putting the first last. Rugby: Practical Action Publishing.
Chang, A.M. (2018) Lean impact. New York: Wiley.
Conway, R., Masters, J. and Thorold, J. (2017) From design thinking to systems change: how to invest
in innovation for social impact. London: Royal Society of Arts.
Cottam, H. (2018) Radical help: how we can remake the relationships between us and revolutionise
the welfare state. London: Virago Press.
Design Council (2019) ‘The design process: what is the double diamond?’. Webpage
(www.designcouncil.org.uk/news-opinion/design-process-what-double-diamond).
DDD Manifesto – Doing Development Differently (2014) ‘The DDD manifesto: on doing development
differently’. Building State Capability, Center for International Development
(https://buildingstatecapability.com/the-ddd-manifesto).
Edmondson, A. (2010) Teaming: how organizations learn, innovate, and compete in the knowledge
economy. San Francisco CA: Jossey Bass.
Freeman, A. (2018) ‘What Skyscanner teaches us about creating a culture of learning through
experimentation’. Blog, 19 July. Brink (https://medium.com/hellobrink-co/what-skyscanner-teaches-
us-about-creating-a-culture-of-learning-through-experimentation-e2e0f093d820).

25
Goeldner Byrne, K., Sparkman, T. and Fowler, B. (2016) Getting there from here: knowledge,
leadership, culture, and rules toward adaptive management in market systems programmes. London:
The BEAM Exchange (https://beamexchange.org/uploads/filer_public/58/52/5852dce7-e660-482c-
aea9-b5613f36f227/adaptive_management.pdf).
Gugelev, A. and Stern, A. (2015) ‘What’s your endgame?’ Stanford Social Innovation Review, Winter
2015 (www.ssireview.org/articles/entry/whats_your_endgame).
Haugh, K. and Salib, M. (2017) Evidence base for collaborating, learning and adapting: summary of
the literature review. Washington DC: USAID.
Heifetz, R.A., Linsky, M. and Grashow, A. (2009) The practice of adaptive leadership: tools and tactics
for changing your organization and the world. Cambridge MA: Harvard Business Press.
HM Treasury (2020) ‘Agile digital and IT projects: clarification of business case guidance’. Webpage
(www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-
governent/agile-digital-and-it-projects-clarification-of-business-case-guidance).
Hogg, S. and Leftwich, A. (2008) The politics of institutional indigenization: leaders, elites and
coalitions in building appropriate and legitimate institutions for sustainable growth and social
development. Development Leadership Program Research Paper 2. Birmingham: DLP.
Hudson, D., Marquette, H. and Waldock, S. (2016) Everyday political analysis. Birmingham: DLP.
IDEO Design Kit (n.d.) ‘What is human-centered design?’. Webpage (www.designkit.org/
human-centered-design).
Interaction Design Foundation (n.d.) ‘User experience (UX) design’. Webpage (www.interaction-design.org/
literature/topics/ux-design).
James, C. (2011) Theory of change review: a report commissioned by Comic Relief. London: Comic Relief.
Jean, I. (2017) Beneficiary feedback: how we hinder and enable good practice. PPA Learning Paper.
London: Bond.
Jeremiah, J. (2017) ‘Survey: is agile the new norm?’. Tech Beacon (https://techbeacon.com/
survey-agile-new-norm).
Knapp, J., Zeratsky, J. and Kowitz, B. (2016) Sprint: how to solve big problems and test new ideas in
just five days. London: Bantam Press.
Ladner, D. (2015) Strategy testing: an innovative approach to monitoring highly flexible aid programs.
Working Politically in Practice Series Case Study No. 3. San Francisco CA: The Asia Foundation.
Laws, E. and Marquette, H. (2018) Thinking and working politically: reviewing the evidence on
the integration of politics into development practice over the past decade. Birmingham: TWP
Community of Practice (www.odi.org/publications/11115-thinking-and-working-politically-
reviewing-evidence-integration-politics-development-practice-over).
Leftwich, A. (2011) Thinking and working politically: what does it mean? Why is it important? And
how do you do it? Birmingham: DLP.
Leftwich, A. (2012) ‘Coalitions in the politics of development: findings, insights and guidance from the
DLP Coalitions Workshop, Sydney, 15–16 February 2012’. Birmingham: DLP.
Meadows, D. (1999) Leverage points: places to intervene in a system. Hartland VT: The Sustainability
Institute (http://donellameadows.org/wp-content/userfiles/Leverage_Points.pdf).
MercyCorps (2017) ‘Agility and evolution: how Mercy Corps adapts at the organizational level’.
Online presentation (https://mercycorps.kumu.io/agility-and-evolution-f8eb7880-144f-4aeb-
be44-d9628593825c).
Muyumbu, G. (2018) ‘The DDD agenda: questions from a development practitioner from the South’.
Care Insights, 15 March (https://insights.careinternational.org.uk/development-blog/the-ddd-
agenda-questions-from-a-development-practitioner-from-the-south).
Natsios, A. (2010) The clash of counter-bureaucracy and development. Washington DC: Center for
Global Development.
Pascale, R., Sternin, J. and Sternin, M. (2010) The power of positive deviance: how unlikely innovators
solve the world’s toughest problems. Cambridge MA: Harvard Business Press.

26
Patton, M.Q. (2010) Developmental evaluation: applying complexity concepts to enhance innovation
and use. New York: Guildford Press.
Rahman, A. (2018) ‘What is frontier technology livestreaming?’. Blog, 14 June. Medium
(https://medium.com/frontier-technology-livestreaming/what-is-frontier-technology-
livestreaming-d0275608750e).
Rahman, A. (2019) ‘Play #5: get to “yes, if” from future investors’. Blog, 11 February. Medium
(https://medium.com/frontier-technology-livestreaming/play-5-get-to-yes-if-from-future-
investors-f95aae259f8e).
Ramalingam, B. (2013) Aid on the edge of chaos: rethinking international cooperation in a complex
world. Oxford: Oxford University Press.
Ramalingam, B., Wild, L. and Buffardi, A.L. (2019) Making adaptive rigour work: principles and
practices for strengthening MEL for adaptive management. Briefing Paper. London: ODI
(www.odi.org/publications/11311-making-adaptive-rigour-work-principles-and-practices-
strengthening-mel-adaptive-management).
Rasmusson, J. (2014) ‘What is agile?’. Webpage (www.agilenutshell.com).
Ries, E. (2011a) The lean startup: how constant innovation creates radically successful businesses.
London: Portfolio Penguin.
Ries, E. (2011b) How Dropbox started as a Minimal Viable Product. TechCrunch, 19 October
(https://techcrunch.com/2011/10/19/dropbox-minimal-viable-product/).
Ries, E. (2017) The startup way: how entrepreneurial management transforms culture and drives
Growth. London: Penguin.
Robinson, C. (2019) ‘A simple (but long) run-through of the Systems Changers programme’. Blog,
7 February. The Point People (https://medium.com/thepointpeople/a-simple-but-long-run-through-
of-the-systems-changers-programme-cfb6a077d682).
Rocha Menocal, A. (2014a) ‘“Parlez-vous politics?” Or why working politically is like learning a
language’. Blog, 8 April. Oxfam (https://oxfamblogs.org/fp2p/parlez-vous-politics-or-why-working-
politically-is-like-learning-a-language).
Rocha Menocal, A. (2014b) Getting real about politics: from thinking politically to working
differently. Report. London: ODI (www.odi.org/publications/8325-getting-real-about-politics-
thinking-politically-working-differently).
Samji, S., Andrews, M., Pritchett, L. and Woolcock, M. (2018) PDIA toolkit: a DIY approach to
solving complex problems. Cambridge MA: Center for International Development
(https://bsc.cid.harvard.edu/PDIAtoolkit).
Sharp, S., Valters, C. and Whitty, B. (2019) ‘How DFID can better manage complexity in development
programming’. ODI Briefing Note. London: ODI (www.odi.org/publications/11315-how-dfid-
can-better-manage-complexity-development-programming).
Sims, P. (2011) Little bets: how breakthrough ideas emerge from small discoveries. London:
Random House.
Sparkman, T. (2015) ‘Thoughts on the use of hypotheses in adaptive management’. BEAM Exchange,
6 January (https://beamexchange.org/community/blogs/2015/1/6/timsparkman/).
SPRING and fuseproject (2019) Scaling social business through design thinking. London: SPRING
(www.springaccelerator.org/wp-content/uploads/2019/04/Spring-Book_031219-webFINAL.pdf).
Tantia, P. (2017) ‘The new science of designing for humans’ Stanford Social Innovation Review, Spring
2017 (https://ssir.org/articles/entry/the_new_science_of_designing_for_humans).
Teskey, G. (2017) Thinking and working politically: are we seeing the emergence of a second
orthodoxy? Governance Working Paper Series. Brisbane: ABT Associates.
Therkildsen, O. (1988) Watering white elephants? Lessons from donor-funded planning and
implementation of water supplies in Tanzania. Uppsala: Scandinavian Institute of African Studies.
TWP Community of Practice – Thinking and Working Politically (2013) ‘What is TWP?’. Webpage
(https://twpcommunity.org/about-us/what-is-twp).

27
UK Government (2019) ‘Agile delivery: how to work in an agile way: principles, tools and
governance’. Government Digital Service (www.gov.uk/service-manual/agile-delivery).
USAID (2018a) ‘Discussion note: adaptive management’. Washington DC: USAID
(https://usaidlearninglab.org/library/discussion-note-adaptive-management).
USAID (2018b) Thinking and working politically through applied political economy analysis.
Washington DC: USAID (www.usaid.gov/documents/1866/thinking-and-working-politically-
through-applied-political-economy-analysis).
Valters, C. and Whitty, B. (2017) The politics of the results agenda in DFID: 1997–2017. Report.
London: ODI (www.odi.org/publications/10902-politics-results-agenda-dfid-1997-2017).
Varhol, P. (2015) To agility and beyond: the history – and legacy – of agile development. TechBeacon
(https://techbeacon.com/app-dev-testing/agility-beyond-history-legacy-agile-development).
Viki, T., Toma, D. and Gons, E. (2017) The corporate startup. Deventer: Vakmedianet.
Wild, L., Booth, D. and Valters, C. (2017) Putting theory into practice: how DFID is doing
development differently. Report. London: ODI (www.odi.org/publications/10729-putting-
theory-practice-how-dfid-doing-development-differently).
World Bank (2019) Strengthening the organization. Annual Report 2018. Washington DC: World
Bank Group (www.worldbank.org/en/about/annual-report/strengthening-the-organization).
Yanguas, P. (2018) Why we lie about aid: development and the messy politics of change. London:
Zed Books.

28
Annex 1  Summary of adaptive
approaches
Table A1  Summary of adaptive approaches
Agile Human-centred design Lean startup and Lean Thinking and working Adaptive management Problem driven iterative
(HCD) impact politically (TWP) adaptation (PDIA)
What problem Lengthy linear (waterfall) Products and services are Figuring out the right The anti-politics machine – Many development Stubbornly low levels of
does it solve? design processes meant created without taking thing to build – the thing international development initiatives fail to address capability of developing
software was out of date by needs and preferences of customers want and will actors imposing technical complexity, promoting country governments
the time it was released and users/communities into pay for – as quickly as fixes to inherently political inappropriate and rigid and the failure of aid
did not meet users’ needs account. Lack of creativity possible – and therefore problems interventions that will have programmes that have
to solve complex problems not failing in conditions of little impact attempted to reform these
extreme uncertainty institutions
Original Software development Products, services, systems Startups, technology, International development, International development Governance (broad sense)
sector manufacturing governance
Roots and Spiral development, Design science (1960s), Lean manufacturing Critical anthropology, Decades of development Management literature,
influences extreme programming, design as way of thinking (Toyota), Six Sigma, Agile political science, political literature (and critique). lived experience of
Scrum, Agile manifesto (1970s), IDEO (1991) economy, complex systems Management literature practitioners, lean, agile,
(2001) science, sociology, history on learning and adaptive HCD, TWP
organisations
Core (1) Individuals and HCD is all about: Lean startup favours: (1) Strong political analysis, (1) Focus on solving local (1) PDIA focuses on
principles interactions over processes (1) Building a deep empathy Experimentation over insight and understanding problems that are debated, solving locally nominated
and tools with the people you’re elaborate planning (2) Detailed appreciation of, defined and refined by local and defined problems in
(2) Working software designing for Customer feedback over and response to, the local people in an ongoing process performance
over comprehensive (2) Generating tons of ideas intuition context (2) Legitimise reform at all (2) It seeks to create an
documentation (3) Building a bunch of Iterative design over (3) Flexibility and levels (political, managerial, ‘authorising environment’
(3) Customer collaboration prototypes traditional ‘big design adaptability in programme social), building ownership for decision-making
over contract negotiation (4) Sharing what you’ve upfront’ development design and implementation and momentum throughout that encourages
(4) Responding to change made with the people you’re (Blank, 2013) (TWP Community of the process ‘positive deviance’ and
over following a plan designing for; and eventually Practice, 2013) (3) Work through local experimentation
The Agile Manifesto (Beck (5) Putting your innovative convenors who mobilise
et al, 2001) new solution out in the world all those with a stake in
(IDEO Design Kit, n.d.) progress
 Agile Human-centred design Lean startup and Lean Thinking and working Adaptive management Problem driven iterative
(HCD) impact politically (TWP) adaptation (PDIA)
Core (4) Blend design and (3) It embeds this
principles implementation through experimentation in
(cont.) rapid cycles of planning, tight feedback loops
action, reflection and that facilitate rapid
revision experiential learning
(5) Manage risks by making (4) It actively engages
‘small bets’, pursuing broad sets of agents to
activities with promise and ensure that reforms are
dropping others viable, legitimate, relevant
(6) Foster real results – real and supportable
solutions to real problems (Andrews et al., 2012)
that have real impact
(DDD Manifesto, 2014)

Key concepts Building incrementally, Prototyping, user Validated learning, pivot, Working with the grain; Locally led problem- Isomorphic mimicry,
sprints, retrospectives experience, co-creation value, growth and impact ‘politically smart’; solving, small bets, authorising space,
hypotheses institutions, interests and convening and brokering positive deviance
ideas
Working Time-boxed iterations Inspiration, ideation, Build, measure, learn, Understanding, testing and Rapid cycles of planning, Problem construction
patterns and (sprints), sprint implementation; double decide. Think big but start learning. Everyday political action, reflection and and deconstruction,
techniques retrospectives and planning, diamond; diverge and small. Innovation portfolio, analysis, political economy revision, portfolio of multiple identify change space,
daily stand-ups, Scrum converge; immersion; MVP analysis (PEA), coalitions experiments, strategy trawl design space, rapid
teams design sprints testing, collective action cycles of action and
learning
Measurement Reflective learning, sprint Validation, insights, user Innovation accounting; Open-ended and process Outcome mapping, rapid Problem solved,
and learning retrospectives, evidenced value validated learning indicators, outcome data collection and course milestones, building
by speed harvesting, Strategy Testing, correction, developmental authorising space
greater use of qualitative evaluation, citizen feedback,
case studies open-ended but countable
metrics
Learn more Agile in a Nutshell: www. Design Kit from IDEO.org: Lean Startup Principles TWP Community Doing Development The Building State
agilenutshell.com www.designkit.org from Eric Ries: http:// of Practice: https:// Differently Capability programme
theleanstartup.com/ twpcommunity.org/ manifesto: https:// at Harvard Center
principles buildingstatecapability.com/ for International
the-ddd-manifesto Development: https://bsc.
cid.harvard.edu/
 Evidence.
Ideas.
Change.

ODI is an independent, global think tank,

working for a sustainable and peaceful
world in which every person thrives. We
harness the power of evidence and ideas
through research and partnership to
confront challenges, develop solutions, and
create change.

ODI
203 Blackfriars Road
London SE1 8NJ

+44 (0)20 7922 0300

[email protected]

odi.org
odi.org/facebook
odi.org/twitter