UX Researchers

Run a usability study
before your next standup.

Real participants test your product while an AI moderator asks the follow-up questions you would. No scheduling. No notetaking. Describe what you want to learn β€” Candor handles recruitment, moderation, and transcription. You get insights, not logistics.

Every evaluation on Candor is completed by a real person. Not an LLM. Not a synthetic label. Human judgment.

The Problem

You spend more time on ops than on research

πŸ“‹

Recruiting takes longer than the research

You spend days finding participants, screening them, scheduling sessions, sending reminders, handling no-shows. By the time you run the study the sprint has moved on.

⏱

Moderation is a bottleneck

You can only run as many sessions as you have moderators and hours in the day. Five sessions takes a full week when you account for scheduling, running, and debriefing. And your best moderator is also your busiest person.

πŸ“‰

Insight delivery is always late

By the time you've transcribed, coded, synthesized, and presented findings, the team has already shipped. Research becomes a retrospective exercise instead of a decision-making input.

Use Cases

From usability tests to discovery research

AI-moderated experience evaluation

Real participants go through an experience β€” browsing your product, interacting with a chatbot, completing a coach session, using a service β€” while an AI moderator asks questions, probes on what landed and what felt off, and adapts follow-ups in real time. You write the interview guide and the AI runs the session. Every session is transcribed with key moments annotated. Run 5 sessions overnight instead of across a week.

Participant view
LIVE SESSION
your-app.com/onboarding
ModeratorTalk me through that interaction β€” what felt natural and what felt off?
ParticipantThe opening felt like a template, but once it got personalβ€”
ModeratorWhat made that shift happen?
AI Moderator
1:47
How you'd run it
$ claude "run a 5-person evaluation of our coach interaction β€” probe on what felt helpful, what felt scripted, and whether they'd come back"
What you get back
Themes across 5 human sessions:
Opening felt scripted (4/5 sessions)
Participants said the first message felt "like
a template." Engagement increased once the coach
asked a personal question.
Advice was helpful but too fast (3/5 sessions)
"I wasn't ready for a recommendation yet, I was
still explaining my situation."
Closing drives return intent (5/5 sessions)
Every participant said the summary at the end
made them want to come back. "It showed the
coach was actually listening."
Transcripts: study/coach-eval/transcripts
Learn more about Voice Interviews β†’

Preference testing on design variants

You have 3 design directions and need to know which one users prefer. Run pairwise comparisons with real users β€” they see two options side by side, pick a winner, and explain why. Get a ranked result with agreement metrics in hours, not days. Works for mockups, copy variants, icon options, anything visual or textual.

Participant view
vs
A
Tie
B

β€œWhich of these two onboarding screens feels easier to get started with?”

How you'd run it
$ claude "compare these 3 onboarding screen variants β€” which do users prefer and why?"
What you get back
Ranked by preference (20 human participants, pairwise):
#1 Variant B β€” "minimal + progressive" 74% win rate
#2 Variant A β€” "full form upfront" 52% win rate
#3 Variant C β€” "wizard with illustrations" 34% win rate
Agreement: 0.82 (strong consensus)
Top reason for B: "I could start using it immediately
without filling out a bunch of fields I don't understand yet."
Learn more about Pairwise Comparison β†’

Concept validation with open-ended feedback

Show real participants a prototype, landing page, or concept description and collect open-ended reactions. Free text responses with optional follow-up from the AI moderator if you want richer signal. Use it to validate a direction before investing engineering time.

Participant view
Free text

β€œAfter looking at this page, what do you think this product does? What's clear and what's confusing?”

How you'd run it
$ claude "show 15 users this landing page and collect their initial reactions β€” what's clear, what's confusing?"
What you get back
Themes across 15 human participants:
Clear: core value prop (12/15)
Most participants accurately described what the product
does after 10 seconds on the page.
Confusing: pricing model (9/15)
"Per-session" pricing wasn't intuitive. Participants
expected per-seat or per-month.
Confusing: "AI-moderated" (7/15)
Participants weren't sure if AI means no human is ever
involved, or if it's AI-assisted with human oversight.
Positive: terminal-first positioning (10/15)
Developers and technical PMs found the CLI angle
refreshing. Non-technical PMs were neutral.
Learn more about Free Text β†’

Discovery interviews β€” understand your user's world

Not every study starts with something to test. Sometimes you need to understand your user's context before you build anything. Run AI-moderated interviews where real participants talk through their daily routines, what tools they use, where they feel supported, and where the gaps are. The AI moderator adapts across sessions β€” by session 5 it's probing the themes that emerged in sessions 1-3, giving you the coverage of a week-long field study in a fraction of the time.

Participant view
LIVE SESSION
ModeratorWalk me through a typical morning β€” what's the first health-related thing you do?
ParticipantI check my glucose app, then usually forget about it until lunch...
ModeratorWhat happens at lunch that brings it back to mind?
AI Moderator
4:22

β€œTell me about your daily routine β€” what tools do you rely on and where do things fall through the cracks?”

How you'd run it
$ claude "run 10 discovery interviews β€” understand how patients manage their daily health routine, what apps they use, and where the gaps are"
What you get back
Themes across 10 human sessions (study-level moderation):
Morning check-in is habitual (8/10)
Most participants check a health app first thing
but don't engage with it again until evening.
"It's a glance, not a conversation."
Midday is the gap (7/10)
Participants described lunch and afternoon as
"when things fall apart." No app fills this
window. Most rely on memory or nothing.
App fatigue is real (9/10)
"I have 4 health apps and I use 1.5 of them."
Participants are resistant to adding another
app unless it replaces something.
Accountability > information (6/10)
"I don't need more data, I need someone to
notice when I slip." β€” strongest signal for
human-in-the-loop features.
Coverage report: 12/15 guide topics explored
Gaps: sleep routines, weekend vs. weekday
patterns β€” suggest 3 more sessions.
Transcripts: study/health-discovery/transcripts
Learn more about Voice Interviews β†’
How This Compares

What changes when you drop the overhead

⚑

Setup time

Traditional tools require project creation, screener surveys, panel selection, and scheduling windows. Candor: one command, participants recruited automatically. Your study is live in minutes, not days.

πŸ€–

Moderation

Traditional moderated research requires a live human moderator for every session. Candor's AI moderator runs sessions in parallel, 24/7, and never forgets to ask the follow-up question.

🎯

Time to insight

The traditional pipeline is sessions, transcription, coding, synthesis, report, presentation. Candor delivers transcripts with themes surfaced automatically. Your job starts at synthesis, not transcription.

Methodology

A note on research rigor

We know UX researchers care deeply about methodology β€” so do we. Candor uses randomized presentation order to prevent position bias, counterbalanced pairwise comparisons, and attention checks to filter disengaged participants. Every study includes inter-rater agreement metrics so you can assess consensus at a glance. This isn't a survey tool with a nice UI β€” it's a research platform with proper methodology baked in.

Your next research round starts here

Your next research round can start in 5 minutes.

$curl -fsSL https://candor.sh | bash
Read the docs β†’
Or talk to us about your use case β†’