Lesson 5 of 5
The cheap tests – Evidence without a research budget
Match the artefact to the conversation, then buy real evidence for free – five seconds for first impressions, a hallway for flows, and silence while they struggle.
Before the tests: know which conversation you’re in. Design work climbs a fidelity ladder – wireframe to mockup to prototype – and each rung exists to trigger a different conversation. A wireframe asks “is the structure right?” – so colour feedback derails it. A mockup asks “does this look and feel right?” – with a trap attached: polished screens look done, so people go quiet on structural doubts that were never actually tested. A prototype asks “does this work in someone’s hands?” – and one that’s approved without being tested on anyone has skipped its own purpose. The high-value habit costs one sentence: name the rung before critiquing. “This is still a wireframe, so I’ll hold the visual feedback.”
Then buy evidence. Two methods, both real, both free.
The 5-second test
Show someone a design for five seconds, take it away, ask what they remember and what the page was for. That’s the whole method – attributed to Christine Perfetti and colleagues at User Interface Engineering, mid-2000s – and it works because first impressions genuinely form within seconds, and a five-second exposure isolates that judgement from the considered second-guessing that kicks in with time. It tests something narrow: does the message land, is the purpose obvious, did the hierarchy steer attention where intended – The Eye’s squint test with a stopwatch instead of a blur. It cannot test flows; five seconds structurally can’t hold a multi-step task.
Take one now, on a screen you haven’t seen:
A five-second test, run properly
A page you haven’t seen will show for exactly five seconds
The hallway test
Grab whoever’s nearby – a colleague from another team, someone in the kitchen – hand them the design and one task, and watch. Deliberately informal, and underpinned by real maths most people invoke without knowing: Nielsen and Landauer’s model has a single test user surfacing roughly 31% of usability problems, with returns collapsing by the fifth user – “you are wasting your time by observing the same findings repeatedly”, as Nielsen puts it. The popularised version (“test 5 people, done”) drops his actual recommendation: run as many small rounds as you can afford, fixing between rounds. Three 5-user rounds catch and repair iteratively; one 15-user round documents a design that never got the chance to improve. One caveat the summaries also drop: the maths assumes comparable users – genuinely distinct groups (hosts and guests, say) need 3–4 per group.
And the discipline that makes or breaks a hallway test: stay silent while they’re stuck. The instinct to jump in and explain destroys the very signal you’re there to collect. The hesitation is the finding.
The PM’s version, stated plainly
A 5-second test is a screen-share and one question: “what do you think this page is for?” – asked before you explain anything. A hallway test is a phone handed over with one task and your mouth shut. Neither needs a lab, a budget or a recruiter, and both beat shipping on internal conviction – which is the actual alternative on most teams.
This page practises it:The demo doesn't describe the method – it runs an actual five-second test on you, on a screen you haven't seen, and scores what your memory kept.