Writing Effective Test Tasks
The difference between useless and actionable feedback often comes down to how you phrase your tasks.
The Anatomy of a Good Task
A good test task has four qualities:
1. Action-Oriented
Starts with a verb that tells the tester what to do.
Strong openers:
- “Find...”
- “Try to...”
- “Complete...”
- “Sign up for...”
- “Navigate to...”
Weak openers:
- “Look at...” (too passive)
- “Think about...” (not testable)
- “Do you like...” (opinion, not behavior)
2. Specific End State
The tester should know when they've finished.
Clear: “Find the pricing page and identify which plan fits a 5-person team”
Unclear: “Explore the settings” (when do they stop?)
3. Realistic Scenario
Frame tasks as something a real user would actually do.
Realistic: “You're trying to invite a colleague to your workspace. Do that.”
Artificial: “Click the button in the top right corner” (you're telling them the answer)
4. Neutral Language
Don't signal what you expect or want.
Neutral: “Try to change your profile photo”
Leading: “Use our easy photo upload feature” (you've told them it should be easy)
Task Templates by Goal
Testing Discoverability
- “Without clicking anything, what do you think this product does?”
- “Where would you go to [accomplish X]?”
- “Find the feature that lets you [do X]”
Testing Task Completion
- “Complete [specific action] from start to finish”
- “Sign up for an account and set up your first [item]”
- “Try to [accomplish goal] using this product”
Testing First Impressions
- “Look at this page for 30 seconds. What stands out?”
- “Based on this page, who do you think this product is for?”
- “What's the first thing you'd click on?”
Testing Comprehension
- “Read this page. What do you think [feature X] does?”
- “After looking at this, explain what this product does to me”
- “What questions do you have after seeing this?”
Testing Recovery
- “You've made a mistake and need to go back. Do that.”
- “Something went wrong. What would you do next?”
- “Try to undo your last action”
Common Task Writing Mistakes
Mistake 1: Giving Away the Answer
Bad: “Click the blue 'Export' button to download your data”
Good: “Export your data as a spreadsheet”
The bad version tests whether they can follow instructions. The good version tests whether they can figure out how to export.
Mistake 2: Testing Multiple Things at Once
Bad: “Sign up, complete onboarding, and invite a teammate”
Good (separate tasks): “Sign up for an account” / “Complete the onboarding process” / “Invite a teammate”
Breaking tasks apart shows you exactly where friction occurs.
Mistake 3: Using Jargon
Bad: “Configure the webhook integration”
Good: “Set up automatic notifications to your Slack channel”
Use language your users would use, not internal product terminology.
Mistake 4: Being Too Vague
Bad: “Try out the dashboard”
Good: “Using the dashboard, find out how many users signed up last week”
Specific tasks yield specific insights.
Mistake 5: Making Tasks Too Long
Bad: “Imagine you're a marketing manager who needs to create a campaign for Q4. First, log into the platform, then navigate to the campaigns section...” (continues for 3 more sentences)
Good: “Create a new email campaign targeting users who haven't purchased recently”
Keep tasks focused. You can have multiple tasks for complex flows.
How Many Tasks Per Test?
The Sweet Spot: 3-5 Tasks
- Why not fewer? You won't learn enough to justify the tester's time.
- Why not more? Fatigue sets in after 15-20 minutes. Quality of commentary degrades. Testers start rushing.
Task Time Estimates
| Task Complexity | Expected Time |
|---|---|
| Simple (find something) | 1-2 minutes |
| Medium (complete a flow) | 2-5 minutes |
| Complex (multi-step goal) | 5-10 minutes |
Aim for 10-20 minutes total test time for async sessions.
Ordering Your Tasks
Start Easy
Begin with a task most testers will succeed at. This builds confidence, gets them comfortable thinking aloud, and produces useful baseline data.
Progress to Complex
Save your hardest or riskiest tasks for the middle.
End with Open-Ended
Finish with a question that lets them share anything you didn't ask: “What would you change?”, “What was the most confusing part?”, “Anything else?”
Example Task Order
- 1. Orientation: “What do you think this product is for?” (easy)
- 2. Core flow: “Complete [main user action]” (medium)
- 3. Risky area: “Try to [thing you're worried about]” (potentially hard)
- 4. Secondary flow: “Now try to [related action]” (medium)
- 5. Debrief: “What would you change?” (open-ended)
Context and Instructions
What to Include Before Tasks
Product context (1-2 sentences)
“This is a project management app for small teams.” Don't over-explain. Let them discover.
Think-aloud reminder
“Please narrate your thoughts as you go. Tell us what you're looking for, what you're clicking, and what you expect to happen.”
Reassurance
“There are no wrong answers. We're testing the product, not you. If something is confusing, that's helpful feedback.”
What NOT to Include
- • Detailed feature explanations
- • Apologies for known bugs
- • Hints about how things should work
- • Your expectations for what they'll find
Examples by Product Type
SaaS / B2B Tool
- “You're evaluating this tool for your team. Explore the pricing page and identify which plan fits a team of 15 people.”
- “Create your first [project/campaign/report].”
- “Invite a colleague to collaborate with you.”
E-commerce
- “Find a [product type] you'd consider buying.”
- “Add it to your cart and proceed through checkout. Stop before entering payment info.”
- “You changed your mind about an item. Remove it from your cart.”
Consumer App
- “Sign up and complete your profile.”
- “Find someone to follow or connect with.”
- “Post your first [content type].”
Developer Tool
- “Install the tool using the documentation.”
- “Run the basic example from the docs.”
- “Integrate with your existing project (or describe how you would).”
Task Quality Checklist
Before launching, verify each task:
- ☐ Starts with an action verb
- ☐ Has a clear end state
- ☐ Uses user language, not product jargon
- ☐ Doesn't give away the answer
- ☐ Can be completed in under 5 minutes
- ☐ Tests one thing, not five