Participants complete tasks on their own, with their screen recorded. You watch back what actually happened rather than what people say happened.
When to use one
When you need to see behaviour rather than hear opinion.
Can people complete signup without help? Where exactly do they stall in the checkout? Is the new navigation clearer than the old one?
People are unreliable narrators of their own behaviour, not because they're dishonest, but because most of it is automatic. Watching removes the guesswork.
The three task types
A study can combine them.
Prototype tasks
Test a design before it's built. Prototypes come from Figma.
Connecting your prototype. Connect your Figma account once, and you can pick prototypes straight from your own files, including private ones. You'll be sent to Figma to approve the connection, and it stays connected for future studies.
You can also paste a link to a prototype that's been shared publicly, without connecting anything. That link has to be publicly shared or it won't attach.
Setting the task. Write what you want the person to try to do, then set what counts as success, reaching a screen, or following a route through the flow. Both allow alternatives: mark several screens as success so reaching any counts, and record more than one valid route. For a route, walk it yourself once and it's recorded as the target.
You can ask up to five follow-up questions once they finish.
The prototype runs inside the session and the participant's screen is recorded, so you see exactly where they went, including the wrong turns.
Website tasks
Test a real, live website. It opens in a browser inside the session, the whole thing is recorded, and participants need nothing installed.
Pick a mode first: this is the choice that shapes everything else:
Complete a task: give them a goal and measure whether they get there. Paste the URL of the page that means success, and each session is scored against it. Matching is forgiving (a trailing slash or tracking parameter won't count as a miss), and you can re-score sessions later if you change your mind about what qualifies.
Explore and react: no goal. They browse and tell you what they think. Nothing is scored. Useful for first impressions and reactions to a redesign.
Asking questions. You can ask questions while they browse, and up to five right after they finish. Follow-ups can be text, single choice, multiple choice, rating or scale. Best for finding where an existing experience breaks down.
Video diary tasks
Participants talk to camera and answer your prompts in their own words. Useful when you want tone, hesitation and body language rather than typed answers.
How it runs. They press record once, and your prompts appear one at a time on screen. They answer, move to the next, and it's captured as a single video. You can also let them upload a separate video per question, which suits people who'd rather record in their own time.
Prompts can be more than text. Each one can show an image, play a video or audio clip, put two options side by side to compare, or send them to a website or prototype and ask them to react.
Options worth knowing. You can ask them to share their screen as well as their camera. You can set a minimum recording length (discourages one-word answers). And you can add up to five follow-up questions after the recording.
Best when you want to hear and see the reaction, not read a summary of it.
Writing tasks that work
This is where these studies succeed or fail.
Give a situation, not instructions. "You need a report of last month's spend for a meeting tomorrow" produces real behaviour. "Click Reports, then select date range" tests reading comprehension.
Don't use your own product's words. If your button says Analytics and your task says Analytics, you've told them where to click. Describe the goal in the participant's language.
One task, one goal. Split anything with an "and" in it.
Say what finishing looks like, so people aren't left wondering whether they're done.
Think aloud
You can ask participants to narrate while they work.
It's worth turning on. Watching someone hesitate tells you they hesitated. Hearing "wait, is this the same as the other one" tells you why: which is the thing you can act on.
It records their voice, so they're asked to agree before starting.
Recording
These sessions record the participant's screen throughout. Video diary tasks also record their camera. Participants are told what will be captured before they begin.
Reading the results
You get recordings, plus data on whether each task was completed, how long it took, and where people went.
Watch two or three sessions end to end before touching the summary. Patterns you'd never think to measure become obvious quickly, and the moment where someone goes quiet and starts hunting is usually worth more than any metric.