Lean Startup’s Build-Measure-Learn loop and the PDSA cycle

In his book, The Lean Startup, Eric Ries argues for ways to expand Lean thinking into the realm of startups, into the realm of huge uncertainty.

At the core of his model lies the Build-Measure-Learn loop, which is the key to genuine experimentation and validated learning through working with customers. But what is the relationship between the Build-Measure-Learn loop and the old Lean stalwart, the Plan-Do-Study-Adjust cycle?

The Build-Measure-Learn loop

The first phase in the Build-Measure-Learn loop is the Build phase, in which the goal is to build a minimum viable product as quickly as possible. This is followed by the Measure phase, where the goal is to determine whether real progress is being made, and finally by the Learn phase, where a decision is made whether to persevere (carry on with the same goals) or pivot (change some aspect of the product strategy).

On the surface, this looks like a fail fast, fail often type of approach that lacks the extensive planning Lean companies typically undergo in their PDSA cycles. However, Ries’s own description of the loop takes an interesting turn:

Although we write the feedback loop as Build-Measure-Learn because the activities happen in that order, our planning really works in the reverse order: we figure out what we need to learn, use innovation accounting to figure out what we need to measure to know if we are gaining validated learning, and then figure out what product we need to build to run that experiment and get that measurement. (The Lean Startup, pages 77-78)

Actually, even though it is not explicitly mentioned in the loop description, there is an implicit phase, Plan! This makes the Build-Measure-Learn loop radically different from a typical fail fast, fail often approach and closer a more traditional Lean approach: as the Measure phase is carefully planned in advance, there is a clear hypothesis to test and clear actions to be taken based on the results. Thus, we are back at the scientific method that is at the core of the PDSA cycle as well.

The Plan-Do-Study-Adjust cycle

Now with the implicit phase of the Build-Measure-Learn loop uncovered, it bears a striking similarity to the PDSA cycle:

Plan. Grasp the current situation, visualize the future state, and plan the implementation and measuring of the results.

Do. Carry out the action plan.

Study. Study the results of the planned measurements.

Adjust. Carry on implementing the improvements in other applicable places, or start over from the planning phase.

Short sidestep into OODA

Curiously enough, even though Ries mentions both Deming and Toyota multiple times, he makes no reference to the PDSA (or PDCA) cycle in his book. Instead, he only mentions John Boyd’s OODA (Observe-Orient-Decide-Act) loop as a source of ideas for the Build-Measure-Learn loop.

This seems strange at first, because the OODA loop is markedly different from the Build-Measure-Learn loop, for example in that it does not require one to execute all the steps before moving to a different part of the loop.

However, the context in which OODA was an influence becomes more clear when OODA is shown in its historical context, that of combat operations. The main premise of the OODA loop is that it is constantly went through during military operations, and the entity that is able to proceed through the loop faster will gain an advantage. This is the key idea! Ries’s argument is that rapid Build-Measure-Learn loops are needed in order for a startup to be successful.

Is there a difference between Build-Measure-Learn and PDSA?

No, not really. Build-Measure-Learn is an application of the PDSA cycle.

It is easy to see, however, why Ries would want to avoid the language of PDSA: PDSA, especially through its renowned application at Toyota, is often seen as a slow-paced process where the absolute correct answer is discovered before any action is taken – the planning phase can easily take longer than all the other phases combined!

Yet, there is nothing in the core of the PDSA loop that would force it to be used in this way. Lean is about doing what works, and context is the key. The Build-Measure-Learn loop is based on the assumption that the real customer needs cannot be known in advance even with vigorous planning. In such an environment, it makes no sense to use the majority of time for planning, when we already know that it is not going to improve our proposal. In the context of extreme uncertainty, Build-Measure-Learn is the logical way to apply PDSA.

It is unfortunate that by distancing his loop from PDSA in name, Ries has inadvertently built a problem in the other direction: because the Plan phase is implicit, Build-Measure-Learn loop can easily seem like a typical fail fast, fail often type of approach, even though it is a rapid application of the scientific method (and PDSA) and not a shotgun approach at all!