Eleven UX research methods, and when each one is wrong.
Generated from the same registry the product runs on, so a method cannot appear here that we do not ship — and cannot ship without appearing here.
Every page names what the method measures, how many people it takes before the interval stops embarrassing you, and where it misleads. Running one costs a link in a channel — there is no panel to buy, and the people who only read the results never need a seat.
Five of the eleven run today.
The rest can be built into a study now — the editor is there, the block is real — and cannot be published, because a participant would have nothing to answer them with. Each page says which it is, at the top, before you have read anything else.
We would rather you found that out here than after recruiting. It is the same rule the product runs on: a study holding a method with no runner refuses to go live.
Where do people go first on a screen?
Misleads when the task wording names the label.
What does the screen communicate instantly?
Recall is not comprehension, and it fails on text-dense designs.
Which direction wins, and why?
Preference is not performance, and order effects are large.
Can people complete the flow?
Prototype fidelity caps every finding it produces.
Pick the adjectives that fit.
Measures reaction, not use — and first impressions of visual style decay fastest of all.
SEQ, SUS, UMUX-Lite or NPS.
One item is noisy alone. Anchor SEQ to a task, never to a session — and never read SUS as a percentage.
Open text, choice, Likert or ranking.
People answer the question you asked, which is rarely the one you meant.
Qualify participants before they enter.
Every finding in the study inherits the screener’s assumptions, and the screener is written before you know anything.
A message with no question.
Collects nothing — so a study that only tells people things has not measured anything.
Which one answers your question
Most research questions have one obviously right method and two that would half answer them. The table is the short version; each page carries the long one.
| You want to know | Method | What you need |
|---|---|---|
| Where do people go first on a screen? | First-click test | one frame |
| What does the screen communicate instantly? | Five-second test | one frame |
| Which direction wins, and why? | Preference test | two to four frames |
| Can people find it in your hierarchy? | Tree test | a hierarchy, as text |
| How do people group your content? | Card sort | a set of cards |
| Can people complete the flow? | Prototype task | a prototype |
| Pick the adjectives that fit. | Desirability test | one frame |
| SEQ, SUS, UMUX-Lite or NPS. | SEQ, SUS, UMUX-Lite and NPS | no design needed |
| Open text, choice, Likert or ranking. | Question blocks | no design needed |
| Qualify participants before they enter. | Screener | no design needed |
| A message with no question. | Info screen | no design needed |