Tests & Results
Configure, run, and read UX tests.
Configuring a test
Every test combines a starting point with a goal and an audience:
- URL: where personas begin. Use the full public URL of the page or flow you care about.
- Task description: a natural-language instruction, e.g. "Sign up for a free account."
- Success criteria: how a persona knows it is done, e.g. "You see a welcome dashboard."
- Cohort: the audience. See Personas & Cohorts.
- Panel size: how many personas to run. Five is a good default; larger panels make repeated friction more obvious.
- Guardrails: block irreversible actions (purchases, signups, newsletters) and supply safe form data.
Test modes
Every new test starts with a mode chooser. The mode shapes both the wizard and how personas behave.
Task
Personas follow your instructions and try to complete a specific flow: checkout, sign-up, search. CrowdRunner reports where they got stuck and why. Best when you have a concrete journey to validate.
Exploratory
Personas browse freely with no fixed goal, surfacing what is confusing, missing, or delightful. Best for marketing pages and first-pass feedback. Task and Exploratory are two framings of the same exploratory engine, one goal-directed and one free-roam.
Chaos
Personas deliberately try to break things: long inputs, emoji, double-click submits, hitting browser-back mid-flow. Counter-personas then re-test each finding to confirm it reproduces. Best for pre-launch QA and edge-case hunting.
UAT
Upload acceptance criteria (or paste them in) and personas execute each one as a structured test case, returning a pass, fail, blocked, or inconclusive verdict with evidence. Best for release sign-off. UAT runs on desktop.
Reading the report
Session timeline
Each persona's run is a step-by-step timeline with screenshots. Click through to see exactly what the persona saw, the action it took, and what it was thinking at each step.
Friction points
CrowdRunner flags moments where personas struggled. Each friction point carries a type (confusing navigation, broken element, unclear copy, error state, dead end, and more), a severity (minor, major, critical), a confidence level, the step and screenshot it came from, and often a think-aloud quote. Friction that several personas hit independently is promoted; friction only one persona saw, while others passed the same screen, is demoted as likely noise.
The Verified tier
High-signal findings are re-checked. An issue marked Verified was reproduced on a second pass; others may be flagged as needing verification. This keeps the headline list to problems you can trust rather than one-off blips.
Blocked vs. real defects
When a persona is stopped by something that is not your product, a CAPTCHA, an unreachable environment, or missing test data, that is bucketed as blocked / inconclusive rather than counted as a UX defect, so environment problems do not inflate your results.
Evidence
Reports ship with evidence you can act on: screenshots pinned to each guaranteed issue, a downloadable fix-list formatted for handing to a coding assistant, and session-replay video of the persona's run.
Synthesis & root causes
The synthesis report opens with an executive summary, then separates issues every persona hit from those that split by profile, lists prioritized recommendations with effort estimates, and clusters related findings under a shared root cause so you fix the source, not the symptoms.
Comparing runs
Compare two completed runs side by side, typically the same task before and after a change, or across two URLs. From the Tests page, open the Comparisons tab, create a comparison, and pick two runs. The report highlights differences in completion, satisfaction, and friction.
Sharing results
Generate a read-only share link for any result to send to teammates or stakeholders. No CrowdRunner account is required to view it.