Make the idea
answerable.
A good experiment tells us more than whether a screen looks finished. It gives us evidence for the next decision.
Frame
Start with a person, a task and a decision. State what we need to learn before picking features.
Make
Build a small end-to-end path. It can be rough at the edges, but its central interaction needs to work.
Test
Check the task in a real browser with real constraints. Separate a successful build from a useful outcome.
Decide
Record what held up, what broke and what remains unknown. Continue only where the evidence supports it.
Say what
is real.
Lab work changes quickly. We label studies as local, live or in review and name the checks behind those labels. A working prototype does not stand in for a dependable service, and a passing technical test does not stand in for user demand.
See current experiments