Small Experiments
What is the cheapest thing that would change your mind?
The idea
A good experiment reduces a specific uncertainty cheaply.
In the real world
A clickable prototype tests comprehension before development.
Going deeper
An experiment's value is the uncertainty it removes divided by what it costs, which means the best test is usually the cheapest one that would still change your mind.
Spending six weeks building to find out whether people understand a concept buys the same answer as a two-day prototype, at twenty times the cost, plus a subtler penalty: sunk work makes a negative result harder to act on. Designing a test starts by naming the single uncertainty and the smallest artefact that would resolve it convincingly — convincingly to you, in advance, so the result cannot be argued away.
Where it stops applying
Some uncertainties genuinely cannot be tested cheaply, particularly around long-run retention or rare events. Forcing a small experiment onto them produces a fast, confident, wrong answer.
Why it matters
The value of a test is the uncertainty it removes per unit of cost, not how thorough it looks.
Try this today
Name the uncertainty, signal, and cutoff.
Test yourself
A team spends six weeks building a feature to find out whether people understand the concept. A clickable prototype would have answered it in two days. What did the six weeks buy?
Show the answer
The same answer at twenty times the cost, plus sunk work that makes acting on a negative result harder. A good experiment isolates one uncertainty and resolves it with the least commitment that would still be convincing.
Learn this in the feed Answering from memory, then again days later, is what makes it stick.