Optimize for learning, not shipping
Shipping feels productive, but learning is what validates the business. Ask what you’re still unsure about, not what to build next.
Every business is an experiment built around a hypothesis: that enough people will pay enough for some benefit from something delivered in some way to sustain it.
The variables in this hypothesis are what differentiate every business:
- enough people: Who the customers are
- pay enough: What & how they pay
- some benefit: What needs of theirs are being addressed
- something delivered: How those needs are being addressed
- in some way: How the business operates to address those needs
The goal, then, of the business is to learn which values for those variables validate the hypothesis. The most successful businesses are those who discover those values as quickly and efficiently as possible.
The learning loop is simple:
- Make a bet about one variable value
- Test that bet
- Evaluate the results
- Repeat
Shipping is an important part of step 2, but many businesses skip the other steps completely. Why? They overestimate their own ability to make good bets, they are too attached to a specific set of bets, and because shipping feels productive.
But shipping without knowing or measuring what you’re trying to learn is like playing poker blindfolded. It’s no wonder half of businesses fail within their first four years (source).
AI has exacerbated this problem. Shipping is easier and faster than ever versus making and evaluating bets, so guess where businesses end up spending more of their time. It’s the classic streetlight effect: doing the less impactful thing because it’s easier.
The solution is to stop asking what to ship next. Instead, ask: Which of those variables still have the greatest uncertainty? What is the cheapest way to reduce that uncertainty? The answer is rarely to build more and often to run a smaller, cheaper test.
For example, a common mistake is to keep adding features to a new product before testing whether anyone will actually pay for the core value proposition. That treats “something delivered” as the uncertain variable when “pay enough” usually is. Shipping more doesn’t reduce that uncertainty. But a pre-order, a waitlist with pricing, or even a handful of sales conversations often will.