Hacker News new | past | comments | ask | show | jobs | submit login

tons! I think about this all the time, since I feel a responsibility to try and talk about lean startup in such a way that prevents misunderstanding.

I would say the two most fatal misapplications are:

1. "up and to the right" disease. here you split-test everything and just do whatever moves the numbers. pretty soon you are selling porn or psychic hotlines.

2. "no vision, no problems" error. It's like trying to do science without a hypothesis. In lean startup we emphasize that people trying to predict the future are often wrong, so it's best to experiment and pivot as you learn. But some people interpret this to mean that the future is unknowable, there's no point in having any kind of vision, and you should just ship something and see what happens. the problem with this plan is you are guaranteed to succeed - at seeing what happens. after-the-fact rationalization will prevent any learning, because if you can't fail you can't learn. having a big expansive vision is really helpful because it provides lots of falsifiable hypotheses for testing.

3. "minimum viable crap" sloppy execution. Some people think MVP means just throw garbage at the wall and see what sticks, especially since the M makes people think lean startup is for doing something small. but the truth is if you're doing something small, you don't need MVP or lean startup. you only need an MVP if you're trying something large. further, part of the MVP process is to learn what customers actually value in terms of quality, so we can build something that they perceive as excellent. shipping crap isn't the goal, and people that go on TechCrunch with garbage and then claim "but it's an MVP!" are doing it wrong. the hard truth is that spending more time "perfecting" a product in the absence of feedback often makes the product worse, not better.

The struggle for me seems to be knowing how much is too little and how much is too much. Do you try to err on the side of doing too little at first?

Well, that's a bit like saying "how much science should I use?" There's not really such a thing as "too much science" more like "bad science" or "too expensive science." The key, to me, is to pick an experiment that is well tuned to your circumstance: the right cost, timeline, hypothesis, and clear action for next steps.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact