Innovation Needs a Kill Switch

I think we have become too comfortable with the idea that experimentation is inherently good.

It isn’t.

Trying something new feels productive. It gives teams energy. It creates something to show. It makes us feel like we are moving.

But movement and progress are not the same thing.

Experimentation without a clear reason is just distraction with a prototype.

The question shouldn't be “What can we try?”

It should be “What are we trying to learn, and why does it matter?”

That distinction becomes even more important when resources are constrained.

When you have unlimited resources, you can afford to explore things that might be interesting.

Most companies don't have that luxury.

If an experiment is taking time away from something else, that time has a cost. So the experiment needs to earn its place.

For me, there are really only two reasons to run one.

It has the potential to create disproportionate impact.

Or it can generate disproportionate learning.

If it does neither, I don't know why we're doing it.

This is where I think the idea of part-time innovation gets interesting.

Yes, experiments can absolutely be part-time.

But they need to be small.

A designer shouldn't need to spend three months building a side project to find out whether an idea has merit.

A team shouldn't need a dedicated squad before they know whether the problem is worth solving.

The first version should be almost embarrassingly small.

One hypothesis. One prototype. One user test. One insight.

Then decide.

Continue.

Change direction.

Or kill it.

The last one is important.

We don't talk enough about killing experiments.

There should be a kill switch.

Before starting something, we should know what evidence would make us stop.

Otherwise, an experiment quietly becomes a project.

The project gets a roadmap.

The roadmap gets a team.

The team gets attached to the idea.

And six months later, we're defending something that should have been killed in week two.

That isn't experimentation.

That's sunk cost.

A good experiment should make it easier to make a decision, not postpone one.

This is also why I don't think experiments should replace research.

Experiments are attractive because they feel fast.

But speed doesn't make an experiment useful if we're testing the wrong problem.

Research gives us the why.

Experiments help us understand the how much.

We need both.

If we don't understand the problem, we can build something very quickly that nobody needs.

And if we understand the problem but never test our assumptions, we can spend months debating something that could have been answered in a week.

The goal isn't to choose between research and experimentation.

It's to use the cheapest method that gives us enough confidence to make the next decision.

That also changes how we think about prioritisation.

Not everything that could be useful is worth doing now.

I like a simple equation:

Impact × Alignment × Effort.

The highest priority opportunities are the ones that matter, are aligned with what we're trying to achieve, and can be explored without consuming disproportionate resources.

Everything else can wait.

And waiting isn't failure.

Sometimes the most disciplined decision you can make is not yet.

This becomes especially important when a company is trying to survive.

There is a romantic idea that difficult periods are when companies need to become more innovative.

I think that's only partly true.

Constraints can create creativity, but they can also create desperation.

When resources are limited, innovation shouldn't mean doing more with less by asking people to work harder.

It should mean making sharper bets.

Survival doesn't require less innovation. It requires less waste around innovation.

Small bets.

Fast feedback.

Clear decisions.

Kill what isn't working.

Double down on what is.

That's the operating system.

The irony is that this can actually create more freedom.

When people know that they don't need to defend an experiment forever, they are more willing to try something.

When failure doesn't create a permanent scar, people take better risks.

When the expectation is not “make this work” but “find out if this is worth working on”, the quality of experimentation changes completely.

The point of an experiment isn't to prove that we were right.

It's to find out whether we were wrong.

And if we were wrong, we should be excited to know that early.

Innovation isn't about having more ideas.

It's about having a system that can separate the ideas worth pursuing from the ideas worth killing.

Because attention is a finite resource.

Time is a finite resource.

People are a finite resource.

So every experiment should answer one question:

What are we going to know after this that we don't know today?

If the answer is meaningful, run it.

If the answer isn't, kill it before it starts.

Innovation needs freedom.

But it also needs discipline.

And sometimes the most innovative thing you can do is stop.