The Lean Startup

Some of the takeaways, by section:

vision

a startup is a human institution designed to create a new product or service
under conditions of extreme uncertainty
The startup's job is to create this new product or service, but will only
succeed if it's able to get through multiple learning cycles to build the
product. It's unlikely that you'll start out with something that will actually
work, so it's important to have a hypothesis with some bold assumptions.

The currency of progress in startups is learning. Validated learning is
demonstrated by improvement's in the startup's core metrics.

1. value hypothesis - whether a product or service really delivers value
2. growth hypothesis - how will new customers discover the product or service
   (viral, for example)

steer

build -> product -> measure -> data -> learn -> ideas -> build

Most startups have big, bold assumptions, which should be viewed as leaps of
faith because the entire venture rests on them being true. If they're not true,
it's pretty embarrassing and you'll probably need to come up with some other
bold assumption.

Lots of stuff on Toyota and genchi gembutsu etc.

(an MVP) is not necessarily the smallest product imaginable, though; it is
simply the fastest way to get through the Build-Measure-Learn feedback loop with
the minimum amount of effort
The entire point is to get through the complete cycle as quickly as possible -
not necessary to waste time in any particular area but emphasize going through
the cycle rapidly.

vanity metrics - not helpful, obvious bullshitting

Cohort analysis - helpful to see if metrics are really progressing in the right
direction (e.g. engagement, paying money)

Good metrics should be:

1. Actionable - clearly shows that A caused B, without having a ton of other
   variables to explain away the problem. Good solution for this is A/B test
   experimentation system.
2. Accessible - the metrics must correlate to behaviors that we understand and
   care about, such as users downloading a product, initiating a chat with
   other customers, etc
3. Auditable - need to be able to validate that the data is correct and matches
   our expectations of reality - if we have reports putting out bad data due to
   some internal process failure that's obviously bad

pivoting

May need to pivot multiple times to survive, lots of glossary terms in the book.

If metrics are in place you can look at metrics before optimization/after
optimization and find out if they're growing in the right direction. If not, you
may need to pivot.

a startup's runway is the number of pivots it can still make
Therefore, reducing time through the Build-Measure-Learn cycle means you get
more rolls to try and figure out what works.

pivot - a special kind of changes designed to test a new fundamental hypothesis
about the product, business model, and engine of growth

• zoom-in pivot - single feature becomes the whole product
• zoom-out pivot - single feature is insufficient, make the product much larger
• customer segment pivot - solves a problem for real customers but not the type
  that they originally planned to serve (e.g. LinkedIn)
• customer need pivot - from building a relationship with customers, the
  startup realizes they are not solving a very important problem, but learns
  about another problem that they can solve
• platform pivot - change from application to platform or vice versa
• business architecture pivot - high margin/low volume to low margin/high
  volume or vice versa
• value capture pivot - changing how value is capture from the product (e.g.
  LinkedIn recruiter)
• engine of growth pivot - viral/sticky/paid growth model, switch model
• channel pivot - change how the product is delivered to customers/distribution
  channel
• technology pivot - switching technologies to do the same thing at superior
  price/performance compared to the existing technology

accelerate

Lots of pretty old stuff here.

Small batch sizes are good, makes it easier to prototype and get more learning
done.

engines of growth

• sticky - keep customers paying, reduce churn. Customers stay once you have
  them. if you're getting more customers than the rate you're losing them,
  you're gonna grow.
• viral - customers do the lion's share of the marketing for you. how awesome.
  viral coefficient needs to be >1, so it'll grow exponentially
• paid - if cost per acquisition is less than the lifetime value of a customer,
  it's gonna grow. increase the revenue you get from the customers or find a
  way to acquire them at lower cost

ask five whys

Basically just ask why five times and at each step in the process come up with
some incremental investment that would help prevent the problem from happening
in the future. Lots of stuff on how it shouldn't result with everyone in the
room blaming each other.

conclusion

not a bad book. I think a lot of the knowledge has been disseminated over the
last decade - this book was published in 2011 after all. I think big insight for
me was framing entrepreneurship as a scientific undertaking and with that came a
lot of interesting ideas of how that might work.