How to build awareness for lean experimentation with marshmallows

Tldr: The marshmallow challenge helps entrepreneurs and intrapreneurs to learn about uncertainty, the importance of experimentation vs. planning, riskiest assumption tests and the impact of team size on experimentation.

My friend Alexis Rondeau and myself are constantly thinking about how to bring the lean startup approach to people in the best possible way. And thanks to Eugen Petersen, who built a couple of startups already, I found out about the marshmallow challenge. We ran this marshmallow challenge at our first „Lean Experiments Berlin“ Meetup and in my lecture on „Lean Entrepreneurship“ at the Technical University Clausthal. The objective of the marshmallow challenge is to create the highest possible freestanding structure (with a marshmallow on top) out of 4 central ingredients in 18 minutes:

1 marshmallow - 20 Spaghetti - 1 meter string - 1 meter tape


A team in the challenge should consist of 4 people . There are different insights from the marshmallow challenge that are beneficial for an entrepreneur, whether you are working on your on startup or in a corporate.

What happens a lot of times, is that the teams are all really motivated and very confident that their team is going to master the challenge and creates the highest freestanding structure. Within the 18 minutes people start planning for some minutes. They then build based on the plan and put the marshmallow at the end on top of the structure. This leads to a surprise. In 7 out of 8 cases the freestanding structure wasn’t built to last for 30 seconds after the 18 minutes were over. When we look at entrepreneurial ventures, it is a common pattern, that much time and money is invested based on an initial plan which doesnt work out in the end. This leads to an huge amount of entrepreneurial waste.

There are four very essential learnings that we identified in our sessions and that we can transfer into the entrepreneurial world.

1. Experimenting trumps planning in circumstances of uncertainty.

There are two types of uncertainty Known-Unknowns and Unknown-Unknowns.

We asked the participants (24 until now) of the marshmallow challenge in advance if they know the texture of the ingredients of the marshmallow challenge. 100% of the participants stated so far, that they know the ingredients of the marshmallow challenge with a confidence of 89%.

The different ingredients — especially the marshmallow — are thus falling into the category of unknown-unknowns. The participants actually thought they know the ingredients, although they didnt know them in the end. Participants might know how the spaghetti and the marshmallow behave in a different context — eating, but this information doesnt really help in the marshmallow challenge.

The uncertainty in the marshmallow challenge is quite high for the participants, especially because it is not recognized. Teams that were performing best in the marshmallow challenge were actually applying an experimental approach. They skipped the long planning period and started to build a freestanding structure relatively quickly.

Entrepreneurs deal with many different forms of uncertainty. Uncertainty can be customer driven, problem-driven, solution-driven or technological driven. The task of a lean entrepreneur is to constantly run experiments to lower the risk by turning Unknowns into Known-Knowns.

2. Notice your riskiest assumption

Another very important learning, that we could illustrate to the participants of the meetup and the lecture is, that it is highly important to experiment with your riskiest assumption first. The riskiest assumption in the marshmallow challenge is the marshmallow itself. If you don’t know whether a spaghetti breaks, when you put the marshmallow on top and if you don’t understand how the spaghetti reacts, it can blow up your freestanding structure at the end. The same is true for any entrepreneurial initiative. If you don’t recognize your riskiest assumption — your activities are under very high risks.

3. Beware of the overconfidence bias

We’ve seen in our runs of the marshmallow challenge, that nearly all teams, see themselves as the winning teams. With a confidence level of close to 100%.

With such a high confidence in your mind, there is a big risk of running into the overconfidence trap. This overconfidence trap is not only existing in the marshmallow challenge, but also in the entrepreneurial world. How many teams of entrepreneurs or intrapreneurs have been very confident about their idea, started to implement and realized soon enough, that their idea doesn’t work out. It is interesting to see, that the teams that actually mastered the marshmallow challenge in our case showed a lower confidence than the other teams. It’s very clear, why this is the case. With a high confidence, that you know the ingredients, that you know how to build a freestanding structure, and that your team is going to be the winning team, it is more difficult to experiment. You are very likely more arguing about the right plan instead of starting to experiment with the riskiest assumptions.

4. Team Size impacts speed of experimentation

In our runs of the marshmallow challenge, we realized, that team size really matters to master the challenge. With a group of 2–3 people it is way easier to create a freestanding structure. You have fewer people that you must interact with and this means fewer people to create a consensus on the next best action. This is also a very important insight for entrepreneurial initiatives. The more people you start you journey with, the more people you need to integrate into the decision-making process on the way.


You need to organize yourself to make something happen. But organizing is built on consensual validation which means that you are validating you own assumption by getting other people’s opinions to get to consensus. But this is fundamentally different from the experimental validation of the lean startup. Within consensual validation you already see power plays occurring between the different team members. These power plays and other consensus-seeking activities drain your time to validate your critical assumptions for real.

Please share the article, if you like it and join us at our next “Lean Experiments Berlin” Meetup.

Lutz Göcke