Less Wrong

The Problems with MVPs

In tech and startups, MVP stands for for “Minimum Viable Product”.

It’s part of the glossary of Startupland that - like many words in the tech vernacular - has become increasingly meaningless through overuse and conflation.

Eric Ries, the founder of the Lean Startup, popularised the concept and defined it as:

“that version of a new product which allows a team to collect the maximum amount of validated learning about customers with the least effort”

So effectively, the original concept was the lightest touch (i.e. the most resource-conscious) product than can support continued learning about the problem space a founder is exploring.

I think that still holds as a useful concept but I’m not sure the execution is playing out in line with the core “point”. A few issues I’ve seen:

  1. Hypotheses aren’t well defined: Often we see founders talking about an MVP as if it’s really just a stripped back version of their full product vision. That’s fine, but the bits that remain must be pointed at validating a specific set of hypotheses about the problem or the ways your target user interacts with the problem, such that it actually provides useful, decision-relevant feedback. If not, it’s just a crappy version of your product. The face value issue with this is that the MVP experiments aren’t providing rich insight. The more concerning underbelly is you don’t have the clarity on what it is you’re trying to prove.
  2. The MVP simply sucks too much: Consumers are ridiculously impatient and unreasonable. We’re used to interacting with beautiful products that work seamlessly - the hacky days where you could put something together with sticky tape and consumers would give it a fair go are mostly behind us. Sure, use tools (especially AI) to reduce the cost/time burden to shipping an MVP, but still put craft into it even if it’s only a thin slice of what you want to build. Your job with an MVP is to make it good enough that the consumer continues to engage and therefore provide good feedback. If they churn out immediately, you’re not learning anything.
  3. The hypotheses are benign: To believe you can build a venture-scale product, you have to believe in something contrarian. If it was obvious, other people would already have done it. Validating that contrarian insight is the most important job. Don’t get caught up in ancillary stuff. The fact someone downloaded your new social app does not mean everyone is looking for an alternative for Facebook (which is a terrible and low-information hypothesis anyway). Demonstrating high engagement with a fundamentally different interaction or format (e.g. disappearing messages, short form video from people you don’t know, etc.) is way more useful and interesting. Stack rack your hypotheses from least to most obvious, and work your way down.
  4. You take away the wrong insights: Founders are passionate, and that’s why I love them. But that passion can blur the objective read of data. If MVP data comes through showing low completion rates on the onboarding funnel and low retention on those that do onboard, it can be tempting to fixate on the less brutal interpretation: “We need to fix the onboarding flows”. That’s true, but what you really need to fix is the core product. Making it easier to access a crappy product doesn’t make the product any better. Use a restaurant as an analogy - making it easier to get a booking solves a short term problem to get more bookings, but if the food tastes like crap you have a bigger issue. The problem comes when the chef (the founder, in this tortured analogy) is the one interpreting the data. You have to remain objective as much as you can, and be willing to be the problem. It’s tougher in the short run for sure, but a much better use of your time, money and emotional energy in the long run.

I’m sure there’s more and I will probably add to and amend this over time, but hope these few thoughts are useful to founders when they think about building their MVP.