Net-new insights
vs. surveys
5x
SOUNDS FAMILIAR?
You have the what. You’re asked for the why.
“The open-ends came back as one-word answers. Again.”
Typed open-ends collapse under respondent fatigue. A conversation doesn’t — people talk 5x more than they type.
“The tracker dropped six points. Leadership wants the why by Friday.”
Your survey can locate the drop. It can’t interrogate it. curly re-contacts the segment and asks — in their words.
“How much of this sample is bots and speeders?”
Attention checks catch the lazy ones. A 15-minute audio-visual conversation is a different bar: you can’t fake it.
LIVE PROBE · AFTER A 6/10 RATING
Every scale point gets a follow-up. Every follow-up becomes coded, analyzable data — mapped back to the rating it explains.
THE CORE OF THE METHOD
Your instrument stays fixed: every participant answers the same main questions, with the same scales and the same stimuli. What adapts is the probing — scripted follow-up logic that digs into each individual answer the way a trained interviewer would.
Comparable
Analyzable
Scalable
5x deeper insights
THE DETAILS YOU’LL BE ASKED ABOUT
Everything below runs automatically and in real time — and every step stays inspectable.
Sampling & field
YOUR SOURCES, YOUR QUOTAS
Recruit from any panel, your own CRM contacts, or a shareable interview link. Quotas and prescreening as usual. Fieldwork runs in parallel: hundreds of audio-visual interviews in a few hours, in almost any language.
Instrument
FIXED SPINE, STANDARDIZED STIMULI
Closed and open questions, ratings, rankings — all question types in one coherent conversation. Stimuli are shown identically to everyone: images, videos, live websites, interactive Figma prototypes.
Probing
CONTROLLED, NOT IMPROVISED
Follow-ups run on defined probe logic that follows interviewing best practices — neutral, non-leading, capped in depth. You review and edit the guide before launch and stay human-in-the-loop during field.
Quality control
YOU CAN’T FAKE A CONVERSATION
Every interview is quality-checked automatically and in real time: consistency, engagement, rushing, fake answers. Voice, tone and non-verbal signals raise the authenticity bar far beyond attention checks.
Flagged interviews are excluded before they touch your data.
Analysis
FROM CODEBOOK TO SIGNIFICANCE
curly generates a codebook, maps variables across interviews, and runs univariate and multivariate analyses with statistical significance testing — live, as the field progresses. Slice by any segment, variable, or hypothesis.
Traceability
EVERY NUMBER HAS A VOICE
Each finding links back to the verbatim quotes and recordings behind it. When someone asks ‘where does this number come from?’, you play them the answer.
QUAL · QUANT · CURLY
Depth
Follow-up questions
Non-verbal signals
Uncovers the 'Why'
Scale
Large Sample Size (n=500+)
Low / no manual effort
Fast turnaround (in days)
Low cost
Output
Qualitative analyses
Quantitative analyses
Statistical confidence
Real quotes
Fraud detection / panel authenticity
METHODS FAQ
If follow-ups differ per person, how is the data comparable?
Can I keep my existing questionnaire and scales?
How does significance testing work on conversational data?
What about bots, speeders and professional respondents?
Where does fieldwork run, and where does the data live?
We’ll run it as a real study — same instrument, hundreds of conversations, full analysis. Judge the method by its output.
Get curly
