Blog
What does it mean to be an Experimental Organisation?
16 June 2026
Experimentation in government tends to get treated as a project: a pilot you run, a trial you commission, an evaluation you publish. Something with a start and an end date. This framing is understandable. It is how most programmes work, how most teams are organised, how most results get reported.
At IGL, we believe it is time to broaden this definition. Experimentation is a practical tool for effective delivery. It connects agility, which is the freedom to act swiftly and adapt on the go overcoming inertia, with systematic learning, which gives you the tools to know if they actually did. Being an experimental organisation goes beyond running one experiment, it is about building the organisational conditions in which testing, learning, and acting on evidence become the default, not the exception.
Agility and learning, and why combining them is harder than it sounds
An experimental organisation normalises this process by creating a structural commitment to closing the gap between guessing and knowing. Part of what makes this valuable is that experimentation converts pure uncertainty into something more workable: rather than having no basis for a decision, organisations begin to understand the probability of what works and under what conditions. And to do so, such an organisation is built on two capacities that pull, at first glance, in opposite directions.
The first is agility: a culture of genuine curiosity, a shortening of the distance between idea and action, and the institutional permission to try things before you are certain they will work. Agility is the capacity to capitalise on opportunities when they arise, to create the conditions in which those opportunities can emerge, and to respond to new challenges and insights without waiting for the approval cycles that most administrations use to manage uncertainty by deferring it. In this sense, agility should not be confused with speed for its own sake.
The second is learning. Learning, in the way we mean it here, is a structured and intentional system for building knowledge: about the nature of the problem itself, its causes, the range of possible solutions, and ultimately what works, what does not, and why; ultimately feeding that knowledge back into decisions. And while the collective wisdom that accumulates in a team over time matters and can be part of it, learning is not informal. It requires time, space, and the conditions in which a negative result can be reported without consequence.
These two capacities create a real tension. Agility pulls toward speed and responsiveness. Learning demands patience and distance from the process. Most organisations resolve this by sacrificing one for the other: moving fast without reflection, or producing learning that is thorough but detached from delivery and never feeds back into what they actually do. Experimentation is what allows you to hold both, not fully resolving the tension but helping reconcile it. A structured test at a contained scale generates fast, credible answers before you commit at national level.
The ten principles of Experimental Organisations
To move this from an ideal to an everyday operating model, organisations need practical reference points. The following principles provide a map for that. They are a set of tools, organised around the three main phases of the policy cycle, that together describe what it means to operate as an experimental organisation, but they are not a linear checklist. An organisation does not need to master all ten at once, but it should be able to locate itself honestly within this framework and have a clear direction of travel.
Phase I – The foundations: Before committing public resources, an experimental organisation establishes a baseline of evidence and clear logic.
LEVERAGE – Use what is known to explore the unknown.
Every intervention should begin with a serious audit of what is already known: what has been tried elsewhere, what the data shows, and where certainty genuinely ends. Respecting existing evidence while admitting the limits of your own knowledge is the starting point of rigour, and the alternative to repeating expensive mistakes that other administrations have already made.
LOGIC – Visualise the path from intervention to impact.
If you cannot draw a Theory of Change, you cannot measure success. Mapping the logic of how and why a programme should work forces teams to make their assumptions explicit, which is the first step to testing them. An untested assumption is just a guess with a budget attached.
IDEATE – Generate a portfolio of policy options.
Complex problems are rarely solved by the first idea on the table, and an experimental organisation cultivates the habit of asking “what if?” before asking “what next?”. Generating a portfolio of competing approaches, including the radical and non-obvious, not just different shades of the same idea, before narrowing the field is not indecisiveness. It is how you arrive at the best option rather than the most convenient one. And narrowing is not permanent: the same openness to re-exploration applies when refining and adapting the chosen approach, reopening the question at each new stage of development (a process described well here).
EXCHANGE – Collaborate to accelerate impact.
Peer exchange, academic partnership, and shared data are core operating tools, and not optional extras. And while we have placed this in the foundations phase, openness is a condition that runs across all three: the organisations that learn fastest are those that treat their doubts, as well as their findings, as things worth sharing with others who are working on the same problems.
Phase II – Test & Learn: This is the core of an experimental organisation, where assumptions are replaced with solid evidence.
TEST – Replace assumptions with evidence.
An untested policy is a gamble with public resources. New programmes should be treated as hypotheses, using pilots, trials, A/B tests, and, where appropriate, randomised controlled trials before any full commitment. An experimental organisation draws on an expansive and blended toolkit of methods, not every question requires a large-scale impact evaluation. What matters is that the design is rigorous enough to generate a credible answer.
LEARN – Place measurement at the core of the organisation.
Evaluation is not something you think about at the end of a programme. It is something you design into the structure from the beginning, linking every intervention to a clear measurement strategy from the outset. What you cannot measure, you cannot manage, and what you measure only at the end, you cannot improve.
DE-RISK – Make small bets to avoid big mistakes.
Spending years and significant public resources on something that does not work, because no one had the conditions to find out first is a failure; unlike discovering it at a small and contained scale. It is also worth being precise about what kind of failure we mean: there is a real difference between poor execution of something well-understood, which is avoidable and should be prevented, and an intelligent failure from a well-designed test exploring genuinely uncertain terrain, which is a valuable data point. It is the latter that experimental organisations treat as a learning asset. A well-intended policy can easily backfire: research has shown, for instance, that providing export information to businesses that were not exporting made them less likely to export. Without testing, we continue to fund programmes based on intuition that can produce the opposite of what we intended. As we argued in our recent piece on moving from aspiration to action, the fear of a negative result is losing its legitimacy as a reason not to test.
Phase III – Act on evidence: Once evidence is gathered, the organisation must act on it. This is, consistently, the hardest phase.
ITERATE – Treat every launch as a new starting point.
The launch of a programme is the beginning of a new learning cycle, and not the conclusion of the previous one. Constant refinement based on real feedback and real data is how programmes evolve from functional to genuinely effective.
STEER – Use evidence and data for strategic decisions.
Scale what the evidence supports. Stop, or fundamentally redesign, what it does not. This requires leaders willing to act with intent on what the evidence says, even when it contradicts what they hoped to find, and even when stopping something means admitting that resources were spent on something that did not deliver. We call this administrative courage. It is rarer than it should be, partly because the incentive structures of most public administrations reward delivery over learning, and completion over correction.
IMPACT – Focus on generating the highest potential benefit.
Data points are proxies for beneficiaries. The measure of an experimental organisation is not the sophistication of its evaluation reports or the number of trials it has run. It is what happens to the welfare, productivity, and life chances of the people its policies are meant to serve. An experimental organisation keeps this in view at every stage: its Theory of Change, its measurement strategy, and its decisions about what to scale or stop are all oriented toward this single question: are we making things better for the people who rely on us?
What it actually takes
There is no single path, and this framework is not a certification programme. But the organisations that have made genuine progress share a few consistent features.
For senior leaders, the work is primarily about permission and tolerance for ambiguity, creating the conditions and opportunities in which experimentation can happen without career risk for those who try things that do not work. The cultural signal that matters most is what happens when a pilot comes back with a negative result.
For technical and operational staff, the shift is more methodological: learning to design programmes as hypotheses, building evaluation in from the start, and reading results with appropriate rigour, including the rigour to resist drawing conclusions from data that is not yet strong enough to support them.
The organisations that have moved furthest almost always have one thing in common: someone inside who believed it was worth trying, and a leadership that gave them enough room to demonstrate it. Whether you are reading this as a senior leader looking to create those conditions, or as a practitioner pushing the agenda from within your unit, the ten principles above are a map. The destination is defined by the impact you are trying to generate.