mark-sweep: in computer science, a simple algorithm for garbage collection.
Hi I'm Dave, I work in product at Meetup.


Double Loop Learning and the Lean Startup

When I was growing up, my dad would try all sorts of crazy management techniques on my friends and I. I distinctly recall being tricked into playing a “game” what was really a simulation of how the People’s Express Airline went bankrupt (true story). My dad was an engineer by trade, but was passionate about management theory. He believed that melding the scientific rigor of engineering with the people science of management would yield great results.

One of the things he subjected me to was the idea of “double-loop" learning. Double-loop learning is a concept in organizational learning theory. Back in the 90’s, the idea of a "learning organization" was a hot buzzword. Chris Argyris and Donald Schön were the originators of the ideas behind the theory. 

Single loop learning is our normal mode of learning. We start with a plan, we act on it and then we check the results. Based on what we find in checking our results, we adjust our next plan and continue through the loop. It looks something like this.

If you’ve read The Lean Startup by Eric Ries, you’ll instantly recognize this image in another form:


This is the startup feedback loop. The iteration model of development. In the Lean Startup, the goal is to minimize the time through this Build-Measure-Learn loop. From out of this concept stems such constructs as Minimum Viable Product, Innovation Accounting and Vanity vs. Actionable Metrics. 

Both of these single loops have their roots in the breakthrough management principles developed by W. Edwards Deming, one of the fathers of the Toyota Lean Manufacturing method. His loop looks like this:

The problem with the single loop cycle is that it has a bias towards efficiency. The goal of the single loop is to do what you’re doing better. Single-loop learning is a feedback loop to correct mistakes and to make production faster. Argyle and Schön argued that what was necessary for organization learning was not just better efficiency, but better efficacy.

Efficiency is making the effect better. Efficacy is being able to do the desired effect. 

In other words, the learning organization has to be able to make sure they are being efficient at the right things. And that requires this learning:

The “double-loop” in double loop learning consists of questioning the basic strategy and foundation on which you are working with. It is taking a look at the assumptions you have, to make sure that you are improving organizational efficacy.

Fast forward to today. The Lean Startup framework also has double-loop learning: It’s called a pivot. When you start off working with the Lean Startup framework you explicitly state your “leap-of-faith" assumptions. These are the baseline hypotheses you are testing with your startup’s initial strategy. 

At a certain point in time, you come back and re-evaluated these “leap-of-faith” assumptions in what is called a pivot or persevere meeting. In this meeting, you decide whether to continue to optimize your current strategy, or pivot to a new strategy/assumptions based on your learning. 

This is double-loop learning baked into your startup. If you decide to persevere - you are continuing to do single-loop learning. You are getting more efficient at your current Build Measure Learn loop. If you decide to pivot, you are engaging that double-loop. You’re going back and questioning your assumptions and developing a new set of assumption to test. This is the heart of Argyle and Schön’s organizational insights.

Eric Ries has said so himself that the Lean Startup is about a revolution in two subjects many engineers never are exposed to: management and accounting. It’s great to see these powerful concepts influencing a new generation of organizations. Looks like we’ve come full circle.