The smallest thing worth shipping

You’ve found two people with a real problem. The tempting next step is to build the product properly: accounts, billing, a dashboard, an app. Weeks later you launch, and only then do you find out whether anyone wants it.

Eric Ries, who popularised the term, defined a minimum viable product 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.” In the same post he warns that an MVP, despite the name, “is not about creating minimal products”. It’s about learning. The question is what you most need to learn.

Find the riskiest assumption

Every idea rests on assumptions. For an app that alerts shops when their site goes down:

• Can we detect an outage? Yes. Uptime checks are well understood. Low risk.
• Can we send an alert? Yes. Email and SMS are solved problems. Low risk.
• Will shop owners pay to hear about outages quickly? Nobody knows. This is the one that sinks the business if it’s wrong.

Build to test the risky one first. Everything that doesn’t help answer it can wait.

1. 1Write down what has to be true for the idea to work.
2. 2Rank them. The riskiest is the one you know least about that would sink you if wrong - here, whether anyone pays.
3. 3Design the smallest experiment that tests that one assumption, and run it with real people.
4. 4The result decides what you build next - or saves you from building at all.

Do it by hand first

You don’t need software to find out if people will pay for alerts. You can check their sites with a script on your own laptop, text them yourself when something breaks, and send a payment link. It looks embarrassingly manual. It also gets you an answer this week instead of next quarter.

Paul Graham’s essay Do Things that Don’t Scale makes the general case. Founders have to “recruit users manually” at the start, and with only a few users “you can sometimes get away with doing by hand things that you plan to automate later.” The essay’s point is that this lets you launch faster.

When the manual version stops working

Doing it by hand is a phase, not a plan. When you’re texting thirty owners a day and they’re paying, that’s the signal to automate the part that hurts most, and you’ll know which part that is because you’ve been doing it yourself.