Lean Startup Distilled
[You can find a German translation here.]
There are quite a few books out there on the topic of Lean Startup plus an endless stream of blog entries and conference talks to keep you busy reading as long as you wish. I have tried to summarize my understanding of Lean Startup in the following 5 methods and accompanying practices. Obviously, they don’t describe everything there is to it, but for me, this is the core of Lean Startup.
Lean Startup presumes that the business model that your product or service is based on consists of a set of assumptions. These need to be validated by experiments and measurements. This approach is based on the scientific method, which essentially consists of forming a hypothesis, performing experiments and analyzing the collected data. The assumptions are collected in a business model canvas, which depicts the business model in an easy and clear manner by splitting it up in few distinct parts.
This canvas is in no way a static document but designed to invalidate wrong assumptions and replacing them with new ones. These changes in strategy are called pivots.
The assumptions in the business model can’t be validated by speculating and arguing in your office. This is expressed by Steve Blank’s pithy statement “Get out of the building!” To verify your assumptions you need to get feedback from real (future) customers.
In the very early stages of product development, the easiest way to do this are interviews. Even though you can only reach very few users in this way you will get indispensable qualitative feedback. A possible structure for these interviews can be found in Ash Maurya’s book Running Lean.
Generally, one tries to (in-)validate one’s assumptions by running targeted experiments. It is important to define specific failure/success criteria before performing the experiment. Otherwise, you will find yourself arguing about what the collected data means most likely trying to justify the result you wanted to see. Confirmation bias is not your friend. Experimenting requires a company culture that allows for failure. The highest potential for learning lies in invalidating assumptions. If invalidated experiments are regarded as failure learning is hampered.
To run experiments as cheap and quick as possible one tries to keep the necessary product increment small. This is what’s called a Minimal Viable Product (MVP). Something minimal that is allowed to be incomplete, but is enough to run the experiment and learn from it. For instance by doing costly calculations, like matching in a dating platform, by hand before automating it (concierge MVP).
Generally, Lean Startup values learning more than the product or growth. This is to figure out what the market really wants before building the product. Learning, rather than the number of implemented product features, therefore, becomes the measure for progress. This can feel highly unsatisfying at first because learning as opposed to implemented requirements is very intangible and hard to measure. The business model canvas is a way to visualise the learning progress.
The learning process is depicted with the build-measure-learn cycle. The cycle consists of three steps. Building something minimal, measuring changed user behaviour and learning from it. An experiment is a full loop through this cycle. This full loop is what you should aim to shorten and optimize for, not any individual step. It doesn’t matter how quick you can build a new feature if you can’t measure changed user behaviour. To optimize for learning one plans an experiment in the opposite direction. First, decide what needs to be learned next, then what you need to measure to do so and finally what needs to be built to make that measurement possible.
As depicted in the build-measure-learn cycle one tries to learn by measurements of actual user behaviour.
A popular set of metrics to measure are the pirate metrics by Dave McClure. They got their name from the acronym of its phases which reads AARRR. These phases form a funnel. In each step, you are losing users. By making changes to your product or service you are trying to reduce the percentage of users you lose in each step.
- Acquisition: How do users find your product or service?
- Activation: Are users executing your core functionality?
- Retention: Are users coming back?
- Revenue: How do you make money?
- Referral: Are users referring you to others?
Do you find that something essential is missing or something else could be left out? Let me know in the comments.
