Photo by Alexandre Debiève on Unsplash

Startups are learning machines. What makes them run?

John Zeratsky


The best early-stage startups are optimized for learning — about their customers, whether their product will work, how to get people to use it, etc.

Oh sure, there’s plenty of other stuff that startups have to do. And like all businesses, they exist to create value for their shareholders, customers, and communities (ideally 🙏).

But none of it matters to a startup if they can’t figure out what’s going to work. And… “figuring out”? Well, that’s just a folksy way of saying “learn.”

There are lots of ways for startups to learn. Here are a few that almost everyone uses:

  • Talking to experts, whether those are industry insiders, other startup founders, successful executives, or investors (although I don’t know if investors always have quite as much expertise as they’d like to think 😉)
  • Interviewing customers, often without showing them a product or prototype. I’m not sure why, but chatting with customers pre-product is considered “customer development” and it’s a high-status activity at startups, while showing a prototype to customers is called “user research” and is foolishly given low status. </rant>
  • Building and shipping an MVP, which has become the default learning tool thanks to The Lean Startup, Y Combinator, and a few other influential voices. This is good because it gives startup teams a reality check; bad because it still takes a looong time to build an MVP and the learning usually happens via data, which is necessary but not sufficient (you need to talk to people!).

There are other, less frequently used techniques, like surveys and fake door tests and data analysis — and probably a million others that I don’t know about.

There’s one new technique that I really like: 1-on-1 onboarding for every new customer. As a founder, you can learn a ton, make your customers feel special, and increase their success with the product. There’s one big challenge, though: You need a real product for this to work.

(AFAIK this approach was popularized in the startup world by Superhuman. Vimcal is doing it now for their new calendar app, which is wonderful. (I’m a total calendar nerd, did you know?) Again, I’m sure lots of people are doing this and have done this that I don’t know about 😇)

And of course, I’ve got a horse in this race: The Design Sprint is a high-bandwidth 5-day process for learning through rapid prototyping and moderated customer testing. I’m biased, but I think it works really well: We’ve run sprints with Slack, 23andMe, Medium, Flatiron Health, Nest, Digit, Blue Bottle Coffee, Foundation Medicine, and tons more startups. (Fun: I got a note from Atoms founder Waqas Ali that they used design sprints in the development of their amazing sneakers.)

But this post is not an ad for the Design Sprint process, I promise 🤓

Mostly, I wrote it to organize my thoughts around how startups learn, and to ask for your input:

  • What did I miss? What other ways do startups (internal or external) figure out whether they are on the right track?
  • What works for you? Last time you were doing something big and new, how did you learn whether it was working?

— JZ



John Zeratsky

Supporting startups with capital and sprints. Co-founder and general partner at Character. Author of Sprint and Make Time. Former partner at GV.