Build a culture of experimentation in 4 steps (start absurdly small)

“We should experiment more.” It's on every product team's wall and in every leadership offsite. And yet most teams either never run a real experiment, or run one, feel good about it, and quietly go back to shipping on opinion. As Harvard Business Review put it in Building a Culture of Experimentation, the whole point is to democratize experimentation - to empower people to make good decisions on their own. The hard part isn't believing in it. It's starting, and then making it spread.

Here's the thing nobody says out loud: you can't mandate a culture. You can't send a memo that says “be data-driven now.” A culture of experimentation is grown, not declared - one small, visible win at a time. This is the 4-step path I use, drawn from the work of David Bland, the person who arguably teaches experimentation better than anyone.

Step 1 - Run one experiment

Not a program. Not a framework rollout. One experiment. Follow the principle of least effort: pick a single risky assumption and run the smallest test that gives you enough evidence to validate it. The goal at this stage isn't rigour - it's momentum. You're proving to yourself, and the people watching, that a scrappy test can settle a debate that would otherwise have gone to whoever argued loudest.

Step 2 - Move to multiple experiments

Turn what you learned into action, then raise the fidelity: run a slightly bigger test that gives stronger evidence. But the real move in step 2 isn't the second experiment - it's the communication around it. Tell the team and the stakeholders exactly how the experiment shaped a decision. Experiments don't spread because you tell people to experiment; they spread because people watch a test change someone's mind and think, “I want that for my decision too.”

Step 3 - Expand to other teams

Now widen the circle. Bring other teams into the process - and get explicit buy-in from stakeholders, which is far easier once you have a couple of wins to point to. Run experiments together across teams and learn from each other. This is where “that PM's thing” starts becoming “how we work.”

Step 4 - Foster it company-wide

Use the momentum. Identify the processes and structures that quietly block experimentation - sign-off chains, output-based targets, roadmaps locked a year out - and change them so that testing is the default, not the exception. This is the step that turns a habit into an operating model.

Why this order matters

Most experimentation initiatives die because they try to start at step 4 - a big top-down “culture change” with no evidence behind it. It's backwards. Evidence earns trust; trust earns scope. Start with one test that settles one real argument, make sure everyone sees how it helped, and let the momentum pull the next step into place.

Every experiment we ran to lift UX Pilot's paid conversion or cut elyps' onboarding churn started exactly here: one risky assumption, one small test, one decision made on evidence instead of opinion. It's also the muscle we build with teams in our Product Discovery Workshop - Test Cards, Learning Cards, and all.