Experiment: Learn your way forward

Most organisations have a complicated relationship with failure. In strategy documents, failure is rebranded as learning. In practice, it is still something to be avoided, explained away, or quietly buried. The result is that organisations keep planning as if enough thinking upfront could remove the need to find out what actually works.

Planning and experimenting are not opposites. But they answer different questions. Planning tells you what you intend. Experimenting tells you what is true.

Experimentation means testing assumptions in conditions that allow you to learn from what happens, then adjusting based on what you find. It is not about running pilots that never lead anywhere, or about novelty dressed up as innovation. It is about treating your current understanding as provisional, and creating conditions safe enough to discover where it falls short.

The key word is safe. A well-designed experiment generates real learning without betting the organisation on an unproven idea. You test something small. You pay attention to what happens. You take that feedback seriously. You adjust. Then you go again.

What makes this hard is not the method. It is the culture. Most organisations, without intending to, punish the honest acknowledgement of uncertainty and reward the appearance of confidence. In that environment, people stop raising what they genuinely do not know. The experiments that happen are the ones expected to succeed, which means the most important questions never get tested at all.

Your organisation’s endocrine system handles feedback loops: the mechanisms that register what is happening and adjust behaviour accordingly. When those loops function well, the organisation reads its environment continuously and corrects course in response to what it finds. When they break down, the organisation keeps operating on assumptions that have stopped being true.

The nervous system is equally relevant. If people at the front line cannot get their observations heard where decisions are made, the feedback loops are severed regardless of what the reporting structure says. The information exists but it just cannot travel.

But there is a third system worth naming here. In the body map we use to read organisational health, the respiratory system handles renewal through exchange: the capacity to draw in fresh perspectives, partner with different thinking and create the breathing space that genuine learning requires. Organisations that fill every hour with execution leave no room for the kind of reflection that experimenting demands. The experiment needs air to run in. Without it, the organisation can go through the motions of testing while never actually allowing itself to find out something new.

Finally, your skin matters. The skin manages your boundary with the outside world: customers, communities, regulators, partners.

The most valuable experiments are almost always those run in genuine contact with the people your work is for. When experiments stay inside the organisation, tested only against internal assumption rather than external reality, they tend to confirm what the organisation already believes.

Physarum polycephalum, commonly known as slime mould, is a single-celled organism with no brain and no central nervous system. It navigates complex environments by sending growth simultaneously in multiple directions, reinforcing the paths that lead toward food and withdrawing from those that do not. Over time, it arrives at extraordinarily efficient routes without having planned any of them.

In 2010, researchers at Hokkaido University placed food sources at positions corresponding to the major cities around Tokyo and observed slime mould explore the space between them. Within 26 hours, it had produced a network that closely replicated the actual layout of the Tokyo rail system, one that had taken human engineers decades to optimise. The organism had no map and no objective. It had only a mechanism for following what was working and letting go of what was not.

The lesson is not to grow in all directions at once. It is that real-time feedback from the actual environment, rather than modelled assumptions about it, is how effective routes are found. The difference between an organisation that experiments well and one that does not often comes down to whether there is a functioning mechanism for feedback, and whether that feedback is genuinely acted on.

In the early 1970s, Yvon Chouinard’s climbing equipment company was the market leader in pitons: the steel spikes hammered into rock faces to anchor safety ropes. They were well-made, in demand, and profitable.

On a climbing trip, Chouinard observed that the repeated hammering and removal of pitons was visibly degrading the rock faces he valued. His own most successful product was damaging the environment that gave his work meaning.

Rather than rationalise this away, he began experimenting with an alternative. Aluminium chocks, borrowed from British climbing practice, could be slotted into natural cracks without hammering, leaving the rock undisturbed. He tested them, refined them, and eventually stopped producing pitons entirely, even though that meant abandoning a major revenue stream. His 1972 catalogue devoted two pages to explaining why, and openly invited customers to change how they climbed. Within two years, the chocks had outsold pitons across the industry.

This is not a story about a pivot or a brand strategy. It is about an experiment rooted in honest observation, and the willingness to act on what it revealed even when the existing approach was still working. The question Chouinard asked was not “how do we grow this product line?” but “is this product doing what we actually want it to do in the world?” That shift in question changed everything that followed.

Harvard Business School professor Amy Edmondson has spent over two decades studying what she calls psychological safety: the shared belief within a team that it is safe to take interpersonal risks, including raising a concern, admitting uncertainty, or reporting a mistake.

Her research consistently shows that teams with higher psychological safety outperform those without it, not because they make fewer mistakes, but because they learn from them faster. In high-stakes environments including healthcare and aviation, the teams with the best safety records are not those where errors are rarest. They are those where errors are most readily surfaced and addressed.

Without psychological safety, experimentation is performance. People try the things they already expect will work, report what confirms existing assumptions, and avoid the honest inquiry that would generate genuinely useful learning.

Be clear about what you are trying to learn, not just what you are trying to achieve. An experiment without a genuine question is just a change. Before you begin, be able to say: we believe that if we do X, then Y will happen, and here is how we will know whether we were right.

Start small. Smaller experiments generate faster feedback, carry less risk, and produce more honest results because less is riding on them. Resist the impulse to make experiments big to justify the effort before you know whether the idea works.

Create conditions where it is safe to find out the answer is no. If only the experiments that confirm existing assumptions are acted on, you are not experimenting. You are validating. The most valuable thing an experiment can tell you is that your assumption was wrong, provided the culture allows that result to be heard.

Stay close to real feedback from real people. What your customers, staff, community and partners are actually experiencing in contact with your work is the only data that matters. The rest is noise dressed up as evidence.

The most common failure mode is the pilot that never becomes a decision. The experiment runs, the results come back, and then nothing happens. Either because the results were inconclusive, or because acting on them would require changing something that powerful people in the organisation are not ready to change.

If your organisation runs experiments without a sincere commitment to acting on what they reveal, you are signalling to the people doing the work that honesty has no practical effect. They will stop being honest.

A second difficulty: experimentation requires tolerance of a temporarily messier picture. When you are really learning (preferably through play), the answer is not yet known. That uncertainty can feel irresponsible to boards, funders and stakeholders who expect confidence.

Managing that expectation is part of the work.

Finally, not everything can or should be treated as an experiment. Some decisions have consequences too significant or too irreversible to test at scale. The skill is in knowing which assumptions warrant a careful small test and which require a different kind of decision-making altogether.

Identify one assumption your team is currently acting on as if it were a fact. It might be about what your customers need, how a process works best, or why a problem keeps recurring.

Then ask: what is the smallest thing we could do in the next two weeks to actually test that assumption? Do that thing, and pay attention to what you discover.

• Where in your organisation are you relying on assumptions that have not been tested in a while? What might have changed in the conditions around them?
• When something does not go as expected, is it treated it as a failure or as information? What does that tell you about your relationship with learning?
• Are there things you have continued doing simply because they used to work, without asking whether they still do?
• What would it mean in your specific context to make it truly safe to find out the answer is no?
• If your organisation’s feedback loops were functioning well, what would they be telling you right now that is currently not reaching the people who need to hear it?

Reframing opens up new ways of seeing. Experiment is how you find out which of those perspectives actually holds up in contact with reality. The next action, Grow, explores what becomes possible when this kind of learning is not occasional but continuous.

Disclaimer: This post was written in collaboration with Claude AI. It was used to help research some of the stories for our examples, and sense-check that what I have written is coherent and tonally consistent.