The Lean Startup

This post is about Eric Ries's book "The Lean Startup". This book's message has a wider applicability than one might think, because contrary to what the title suggests, its methodology applies to more than just startups. There's a telling phase that Eric uses to define a startup, which is “a human institution designed to create a new product or service under conditions of extreme uncertainty”. I love this definition because it says nothing about what the institution looks like or how big it is. In fact, lean startup principles very often get applied to small "innovation teams" within large, otherwise slow-moving and highly bureaucratic companies. Reis talks about this in the look, stating that there are basically 3 things needed for innovation to thrive - scarce but secure resources, independent authority to develop the business, and a personal stake in the outcome. Teams with these characteristics can deliver surprisingly powerful results.

There are a couple key ideas to the lean startup approach that are worth discussing. The first is a concept Reis calls "validated learning". I think of validated learning as applying the scientific method to building a business. The idea is to systematically test your assumptions by treating everything as an experiment. Specifically, you should set up experiments for every assumption or decision in such a way that there is a non-ambiguous measurable component (defined a priori) that tells you if you were right or not. Every product, feature, marketing campaign etc. becomes an experiment. Done correctly, these experiments result in empirical demonstrations that valuable truths about the business’s prospects have been discovered.

This idea seems to mesh pretty well with what Nassim Taleb calls "tinkering", which is experimentation without necessarily having a clear thesis about what you hope to find. Tinkering is an asymmetric process because the upside of discovering something can be many orders of magnitude greater than the cost of doing the experiment. In Reis's model the experimentation is more strongly guided by prior assumptions, but the asymmetry still holds. I also think adopting this philosophy would lead to making predictions that are more easily testable (almost out of necessity), which is a good way to calibrate prior beliefs on a wide range of topics.

A related thread in the book is the idea of only doing work that's necessary to achieve validated learning. Anything that doesn't contribute to learning is a form of waste, so just do the stuff that's absolutely necessary to test your assumptions. I imagine this is harder to put into practice than it sounds. It's not at all trivial to identify what's really adding value when you're trying to make decisions day-to-day about what to focus on. There's probably a lot of gray areas and I doubt anyone actually, literally tracks 100% of their time and effort according to this metric, but it seems like a good heuristic.

Another key idea from the book is the concept of value vs. growth hypotheses. It goes like this - the two most important assumptions to make for a product are its value hypothesis (does it deliver value to customers once they're using it) and its growth hypothesis (how will new customers discover it). Every assumption falls into one of these categories. The value hypothesis must be proved before the growth hypothesis. If you've adopted the "validated learning" philosophy, this means that experiments testing the value hypothesis are essentially an early version of your product.

I like this distinction because it has a focusing effect when deciding what to work on. There's no point worrying about how you're going to grow if you don't have a product that anyone wants to use. This directly ties into two of the most famous (infamous?) concepts from the book, the minimum viable product (MVP) and "pivoting". An MVP is the smallest set of features that can be put together to test the value hypothesis. It's a bare-bones version of the product you hope to build. If the MVP doesn't work (i.e. if customers do not find the value hypothesis compelling) then it's time to pivot. Pivoting is just testing a new value hypothesis. It's adjusting your business assumptions in the face of new evidence gained from testing the MVP.

Both of these ideas have permeated popular culture (or at least startup culture) to the point where they're way overused. I think Reis's original intent with MVPs and pivoting makes a lot of sense, but there are some valid criticisms. The biggest one is just how subjective all of it can be. It's not like you get a simple binary pass/fail. The initial product might be kind-of sort-of working, but not quite working well enough, but maybe it will with a few more tweaks or by adding features X and Y. It's very hard to know where you're really at, and no amount of measuring and experimentation can eliminate that ambiguity. Still, as a framework for getting to a viable business model it's a very logical approach.

Overall, I thought this was a great read. Reis's framework for product development turns out to be surprisingly flexible. Even though the focus of the book is on building new businesses, I think the concepts are general enough that they can be applied to a much wider set of circumstances. In a sense, Reis is just broadening the definition of what it means to be innovative (a topic I've written about before) and defining a strategy to consistently achieve innovative results. The book covers a lot more ground than what I discussed here, but it's very accessible and easy to get through in a few days. Highly recommend it.