Eric Ries’ The Lean Startup is a very successful book—and an even more successful movement—amongst product developers. Even if those in product development haven’t read the book, they talk about concepts introduced and popularized in it, such as the ‘Build-Measure-Learn’ cycle and Minimum Viable Products (MVPs).
When defining the ‘Build-Measure-Learn’ cycle, Ries explains that “the fundamental activity of a startup is to turn ideas into products, measure how customers respond, and then learn whether to pivot or persevere.” An MVP, he says, is “that version of the product that enables a full turn of the Build-Measure-Learn loop with a minimum amount of effort and the least amount of development time.”
Whenever I work with a new team, I inevitably hear the phrase “let’s just build an MVP”. Teams resort to this strategy when they are given work that either has a lot of unknowns, doesn’t clearly solve a user problem, or doesn’t align with the goals of their company. What these teams often mean, though, is “let’s build whatever we can build in the time frame that someone has set for us (or that we have set for ourselves) that gets us closer to our grand vision.” The MVP often ends up being a small Version One.
Others in product development have noticed this too, and have proposed a variety of alternatives to ‘MVP’ as a way to capture the nuance of Ries’ definition of a product. Minimum Valuable Product, Minimum Viable Experiment, and Riskiest Assumption Tests are all alternatives that have been proposed in the last couple of years. In his post Death of the Minimum Viable Product!, Steve Cohn, CEO of Validately, shows that the use of the word ‘product’ caused confusion amongst product people because they interpreted it as a ‘releasable product’. His analysis is spot on.
If you look past Ries’ definition of an MVP in The Lean Startup, and instead focus on the examples he uses, it becomes clear that his definition of MVP means something that you can use to run an experiment. Ries gives examples of an MVP, my favorite of which involves Food on the Table. Food on the Table’s product helps users plan meals by comparing their recipes to items on sale at their local grocery stores. The founders of Food on the Table experimented with solutions with customers for months without ever building software. They compared local grocery store sales and meal preferences for each individual customer, printed out recipes and meal plans, and, at first, went shopping with each customer until they had so many customers that the team had to build software to keep up with demand.
It’s fair to say that the printed-out paper the founders provided was, in fact, a product. The problem is that software—or even hardware—developers don’t tend to create and keep version numbers when they move from paper or physical prototypes to software. Calling the printed-out paper that the founders provided to their customers in the beginning a ‘version’ of their product is a lovely idea. But when Ries wrote the definition of MVP, he didn’t mean for ‘version’ to span the entire development process from “I have an idea” to a marketable, sellable product. Because of this, the use of the word ‘version’ didn’t fit with the definition of the word that readers typically understand.
How, then, can we rewrite the definition to capture the nuance that Ries illustrates through examples?
Since an MVP is something that can be used to test out a hypothesis, we could start by saying that rather than an MVP being a “version of the product”, it’s a “testable solution to a problem.” Understanding an MVP in this way frees us up to imagine all of the potential mediums that a solution might be, such as paper, digital, or physical representations.
Let’s move on to the next part of Ries’ definition: “that enables a full turn of the Build-Measure-Learn loop”. Building could, again, involve any medium. So let’s focus on ‘Measure’. To measure something in a meaningful way, a measurement needs to follow the SMART metric methodology: Specific, Measurable, Actionable, Realistic and Time-Bound. Whatever is being tested must have an objective right and wrong answer in a defined period of time. That way, it’s possible to make a decision easily about whether to continue with or change the course of your product. If a test doesn’t have a metric that is SMART, it’s too easy for teams to not learn something from the test and instead succumb to the sunk-cost fallacy, in which they continue to build simply because they have put time and effort into the product. I’d therefore propose that this section of the definition be changed to “used in an experiment that can generate an objective right or wrong answer in a given amount of time”.
The end of the definition, “with a minimum amount of effort and the least amount of development time”, follows standard lean practices. There’s nothing inherently wrong with or confusing about the sentence. But for the clarity of the entire definition, let’s shorten it to “with the least amount of time and development waste.”
When we pull it all together, we arrive at a definition of MVP that is much clearer and captures the nuance of what Ries was trying to teach: “The MVP is a testable solution to a problem used in an experiment that can generate an objective right or wrong answer in a given amount of time with the least amount of time and development waste.”
By being more specific in the definition, I’m hopeful that teams will not resort to saying “let’s just build an MVP” when they are caught in situation in which they are asked to build something without understanding which problem they are solving. Rather, they will help their stakeholders and teammates discover user problems, and then use their MVPs to validate if they are indeed worth solving.
Originally written for the blog on L4 Digital.com.