The Lean Startup

How Today's Entrepreneurs Use Continuous Innovation to Create Radically Successful Businesses

finished (2020)Eric Ries2011


1. Start

In this introduction to the book, Ries lays out the problem: the modern economy has a vast excess of manufacturing capacity, so the critical challenge is to guide it to productive and successful ventures. However, these are hard to come by: it's essentially impossible to make a plan to do something new and be able to follow that through to execution. So, drawing on the Toyota concept of "lean manufacturing", Ries proposes that the most important skillset for startups is the ability to "steer", using feedback loops like the driver of a car rather than trying to pre-program a flight path (as in rocketry). Planning and execution are both important; they should be linked, with relentless usage of the scientific method, so that no precious insights are lost.

2. Define

What is a "startup"—when and where is the advice in this book applicable? "A startup is a human institution designed to create a new product or service under conditions of extreme uncertainty." This ranges from the classic Silicon Valley garage to "intrapreneurs" within established institutions all the way to governmental organizations. It's about the "how" of innovation, not the "where" (which should exist as a prerequisite for applying this book). Specifically, it focuses strongly on scientific experimentation: if one idea gets built, a culture of politicians and salespeople develops; if as many good ideas get built and evaluated as possible, entrepreneurs emerge!

3. Learn

The core goal of a startup should be "validated learning": not post-hoc rationalization of failures as "learning", but actively seeking out what customers are interested in and building features based on that. Make things people want! Any time spent building things that do not improve the experience or give value to the customer is wasted time, no matter how efficiently or beautifully code is built. Entrepreneurship is like striding into the void; follow the firm path of interest wherever it leads, starting as soon as you can possibly begin getting feedback!

4. Experiment

"Planning is a tool that only works in the presence of a long and stable operational history." Every product should be an experiment, with a hypothesis, data, and a conclusion; does it work? Is it worthwhile to build? In today's economy, anything can be built—the question is whether it's worthwhile. Any startup should ONLY be building things that answer questions; don't make assumptions about customers! A startup itself is an experiment, so every product and action it makes should be too.


5. Leap

Absolutely critical to a business's success are two core assumptions: how they think they'll create value, and why they think they'll grow. Together, these "value" and "growth hypotheses" serve to explain how customers—and a lot of them—will come to use the business. At the beginning, however, they are ONLY ASSUMPTIONS. The absolute priority of any startup is to validate these assumptions: if they're wrong, and you don't discern that, you're dead. NEVER BUILD A FULL PRODUCT WITHOUT VALIDATING CORE ASSUMPTIONS FIRST! (But also, don't over-analyze without ever building anything.)

6. Test

How do you validate assumptions? The scientific method. How do you perform an experiment? By building a "minimum viable product" (MVP) which can yield the data you need. MVPS ARE FOR LEARNING, NOT FOR BEING USABLE OR BEAUTIFUL! If you assume that 10% of people will sign up, make a fake ad and prove it. If you think an interface will be usable, scribble it on paper and have them "use" it while you swap out screens. If you think people will pay $10 for a manual service, then charge $10 for that service even if it requires the CEO doing labor and costs $100 out of your pocket. VALIDATING ASSUMPTIONS—THROUGH SCIENCE—IS THE CRITICAL ENGINE OF DEVELOPING NEW THINGS. IF YOU DON'T DO IT, YOUR ENTERPRISE DIES.

7. Measure

To test assumptions, collecting data is critical. Therefore, integrate data collection into everything the product development team does: it's extremely inefficient to have completely separate systems for aggregating data instead of just doing it in the main application/database. (What to measure? Why not everything you can?!) All features should be able to have their changes analyzed, and be revoked if found insignificant or negative (git hygiene!); no feature should be added without running a brief A/B "split test" experiment to see if it improves actual metrics. Metrics-wise: use split tests or cohort analysis instead of gross metrics, which make you feel good but don't actually give good insights!

8. Pivot (or Persevere)

What happens if your tests come back negative? What do you do if a series of attempted improvements fails to have an effect? What comes of being stuck in a hole, unable to get customers to pay for your product (or even sign up!) at the rates you expected? You can "pivot": making a fundamental change to your hypotheses and strategy that still "keeps one foot in the old model," leveraging advantages and pre-existing technology & users from your failed attempt. See what DOES work about your platform—and focus on it! (Or seek out new ways, listen to inquiries from other potentially-interested parties, etc.)


9. Batch

To accelerate the experimentation cycle, it can be helpful to rely on small, fast batches. Instead of taking months to plan out every detail and have one giant launch / product rollout, release changes as you go; use 3D printing and heavy software reliance to be able to prototype things in a time on the order of days or weeks, not months or years. Small batch sizes allow for flexibility, both in innovation and in meeting demand: low "WIP" stock allows for less waste. (This is the "Lean" in Toyota's "Lean Manufacturing"!)

10. Grow

How do startups grow? There are three main methods: the viral, sticky, and paid engines of growth. In viral growth, businesses rely on each of their new customers to bring in more than one extra customer with them (the "viral coefficient" must >1, but should be as big as possible). In sticky growth, you focus on the growth rate (% of new users / total users in any given time) and the retention/dropoff rate. If you grow faster than you wither, you're doing well! The third model, paid growth, focuses on direct/single transactions, and calculates the "cost of acquisition" vs "lifetime value" of a customer. In each case, note that NEW CUSTOMERS COME FROM THE ACTIONS OF PAST CUSTOMERS (directly or indirectly (through $$)).

11. Adapt

"Chronic problems are caused by bad process, not bad people... remedy them accordingly." Institutional problems are inevitable: when big problems arise, aim to understand their root causes and fix them by the "Five Whys" analysis: find the failure at five successively deeper levels of causation, and apply a patch at each. (Surprising fact: almost all (even technical) problems are eventually rooted in human processes!) Short of the ability to do a fair analysis of that—with all stakeholders present to share their perspective—treat first mistakes as learning opportunities. A first mistake is okay; a repeat should never be acceptable!

12. Innovate

Three things are crucial, to nurture disruptive innovation:

  1. Scarce but secure resources: the startup team should have to stretch a budget, but not stress over protecting that budget in the fear that it may suddenly disappear.
  2. Independent development authority: startups should have autonomy, and not be accountable to corporate politics in order to get things done.
  3. A personal stake in the outcome: creating new things is HARD, and people are most motivated to the task when they share in the upside (financial or even repetitional)! Within larger organizations, organizers often take care to "protect" the startup from the legacy business. They should do the opposite: startups are very prone to cannibalize existing markets, and must be shielded from politics in that eventuality! Innovation's pointless when it's not safe to succeed.

13. Epilogue: Waste Not

We should avoid waste—but waste is not primarily the doing of things inefficiently. It's the doing (even efficiently) of the wrong things! The scientific method, applied to technology, has brought us far: today's scientists and engineers can do things past generations could only have dreamed of. If that method is applied even further, to the system in which such wizards work, imagine how much more could be accomplished with the boundless economic capability of humankind!

if this is your website, log in to make changes