The Lean Startup
← Back to books • Amazon • Audible
What’s a startup? It’s a venture in the environment of extreme uncertainty. It’s when you’re not even exactly sure what you should build. The ultimate goal of a startup is therefore not to generate revenue or attract as much users as possible, but to learn what to build and how to create a sustainable business. Therefore, a startup should be patient for growth, patient for revenue, but impatient for validated learning.
Validated learning is a scientific approach to building a startup. It’s about making hypotheses, figuring out the right metrics that indicate success or failure and conducting experiments to test those hypotheses.
Step one is to create a MVP. MVP is a hypothesis for what the product should be. It should omit all features (and the level of polish normally expected) that are not absolutely crucial for testing the hypothesis.
Step two is figuring out the right business metrics. It’s crucial to make a distinction between valuable and vanity metrics. Vanity metrics are the usual, traditional metrics. It’s the revenue, the profit, the number of users, etc. Those are important in a profitable business, but don’t say anything productive in the context of a startup. It’s easy to make the vanity metrics go “up and to the right”. You can call your press friends and get a ton of people (who won’t buy anything). You can spend a lot of money on advertisements and make large revenue (but still lose money). You can make some profit, but still have no clue into why you made it. Instead of tracking people and money in absolute terms, you have to do it in relative terms. You have to track your funnel and such.
Step three. Optimize. Knowing the right metrics, make hypotheses as for what could improve those metrics and then conduct experiments to verify them. Doing tests is “inefficient”, but the ultimate waste, the ultimate inefficiency is creating something nobody wants. Each experiment brings something new to your understanding of what people want.
Small batches. Just like Toyota designed its manufacturing machines for easy retooling, design your tools and processes so that you can release often. Don’t make annual or monthly releases. Release as often as practically possible. (Obviously easier if you’re a website than an app).
Build-measure-learn loop. The batches (the MVP and subsequent releases) should be designed to take you through the iteration of the loop as quickly as possible. The loop actually works in reverse: first you need to figure out what you need to learn, then you need to find the right metrics to measure it and only then do you get to build the right experiment.
Step four. Persevere or pivot. We often hear stories from entrepreneurs that their businesses succeeded because they persevered when people doubted in them. In reality, startups often fail when their founders persevere for too long when the product clearly doesn’t work. If you build an MVP and try lots of experiments to optimize it and it still doesn’t work, you need to pivot. Pivoting means to change your course with one foot on the ground. It doesn’t mean to throw out everything you’ve build and start from scratch, it means to readjust the hypothesis of what the product should be. Sometimes a feature turns out to be the real product, sometimes the MVP turns out to be just one feature of the real product. Sometimes the product is fine but isn’t marketed to the right people.
You need to figure out two things: product value and growth engine. Product value is about what makes people willing to pay, and growth engine is about how to get more people on board. You need hypotheses and experiments for both.
← Back to books • Amazon • Audible
- release often
- test and experiment a lot
- valuable metrics and vanity metrics
- figure out things you need to learn, then make hypotheses and test them