The Culture of Experimentation: Lessons from Booking

Introduction

Throughout my career, one central theme has guided me: when I join a client or project, I want to understand four things as quickly as possible (while respecting any legal constraints):

1. What are we building?
2. Why are we building it?
3. When are we planning to execute?
4. How are we building it?

For a long time, I viewed the “how” primarily in terms of technology, the stack, the architecture and design patterns. I separated it in my mind from the “what” and “why.” But my experience at Booking changed that perspective entirely.

When I started, I expected the usual setup: project managers with KPIs, teams aligning to business goals, and the familiar cycle I’d seen before. But to my surprise, the reality was very different, and far more data-driven than I had imagined.

If the Project Manager Doesn’t Decide, Who Does?

At Booking, decisions aren’t driven by management hierarchy, they’re driven by users.

For the first time in my consulting career, I’ve had access to user interaction data in real time and sometimes hourly. Every click, scroll, page view, action, and even inaction is tracked and fed into a platform called the Experimentation Dashboard.

This dashboard isn’t just another tool. It’s the single source of truth. Whether you’re a developer, a UX designer, a project manager, or a department head overseeing thousands, you see the same data. And here’s the key: nobody can override it.

That reality forces everyone to focus on three things:

1. Generating strong hypotheses (more on this later)
2. Facilitating and refining them
3. Executing them

Life Under the Hypothesis Regime

So what is a hypothesis in this context? Simply put, it’s an idea (an assumption about user needs or behavior) that can come from anyone. But it must be supported by research, historical data, and a clear value case.

For example, imagine you’re working in finance. While planning a trip, you notice that Booking doesn’t properly address event-driven travel (e.g., people traveling to Japan specifically for the Obon Festival). That observation alone could spark a hypothesis.

You draft a short proposal and submit it to the relevant team. From there, it’s not about one person deciding. it’s about the team validating it against data and user research. If the hypothesis holds, it might be live on the platform within weeks.

This openness, where ideas can originate anywhere, creates a culture of shared responsibility and rapid validation.

Executing Experiments

Execution at Booking is fast, pragmatic, and sustainable. The goal is always to find the shortest path technically and legally, to validate or invalidate a hypothesis.

As one of my managers put it:

“Nine out of ten hypotheses won’t see the second month on production.”

That’s by design. Hypotheses are short-lived, highly specific, and judged solely on their impact. Measured in user transactions or satisfaction.

Does This Fast Pace Create Technical Challenges?

Yes and no.

As a developer, you’re equipped with tools, data, and a highly collaborative environment. You’re not just coding, you’re analyzing user behavior, questioning research, voting on priorities, and even weighing in on legal implications.

Technically, this shows up in the codebase through feature flags. Everything is built, tested, and managed with the expectation that it may be short-lived. And when a hypothesis fails, the team is responsible for cleaning up and leaving the system better than they found it.

What Makes a Good Hypothesis?

A few guiding principles stand out:

1. Strive for minimal impact. Small changes (e.g., updating copy on a page) can provide clearer insights than sweeping redesigns.
2. Respect the ecosystem. At any given time, there may be 1,000 active experiments. Your experiment must not conflict with or invalidate others.
3. Always ground ideas in data. No shortcuts. every hypothesis must be backed by evidence.

Conclusion

Experimentation isn’t a phase in the process, it is the process. By empowering teams to test, learn, and adapt based on real user behavior, Booking has built more than a product engine; it has built a learning system.

This culture of experimentation transforms uncertainty into insight, and decisions into data-backed progress. It’s a model any organization striving for sustained innovation should study closely in my opinion.