How Candor speeds up customer research, end to end.

Candor starts with your research question, retrieves real evidence, builds synthetic users grounded in that evidence, lets you interview them, and synthesizes everything into a structured report. Whether you’re running problem discovery, concept testing, price testing, or value-prop validation, the pipeline below is the same. Here’s each step.

STEP 1

Define your audience

Describe who you want to understand, in plain language. Include demographics, behaviors, industry context, and whether you’re researching B2B or B2C. Then choose your study type. Candor supports problem discovery, problem validation, concept testing, and price testing as primary interview types, with value-prop testing and assumption validation layered on top. Each shapes how findings get synthesized. See the full set on the use cases page.

You can also upload your own research: customer interviews, survey data, market reports (PDF, DOCX, CSV, or TXT). Candor parses, chunks, and indexes these documents. They become the primary evidence source for everything that follows, prioritized over web research.

STEP 2

Evidence retrieval and signal extraction

Candor searches your uploaded documents and the web for evidence about your target audience. It extracts hundreds of behavioral signals (behaviors, pain points, attitudes, constraints, goals, beliefs, preferences, and decision rules), each tagged with provenance. That means grounded in your research, inferred from behavioral patterns, calibrated from validated distributions, or flagged as low confidence.

An iterative gap-filling process identifies what evidence is missing and runs targeted searches to fill coverage holes. A critic reviews the evidence for quality and diversity before moving on.

STEP 3

Audience segmentation

Most tools segment by demographics alone. Candor segments by behavioral and psychographic dimensions, the ones that actually explain why people make different choices. Each segment has a primary differentiating dimension, inclusion and exclusion criteria, and a population weight. You review segments and choose which to include before persona generation begins.

STEP 4

Persona generation

From your segments, Candor generates archetypes: the distinct behavioral clusters within your audience. A critic validates that archetypes are truly distinct and evidence-backed. Then it samples individual personas within each archetype.

Each persona gets Big Five (OCEAN) personality traits sampled from peer-reviewed population distributions, calibrated by region and occupation. Cognitive biases are assigned with research-backed intensity values. No two personas in the same archetype are clones. Sibling avoidance ensures each has distinct trait combinations.

Finally, each persona receives rich narrative memory (identity, behaviors, beliefs, communication style) plus seed decisions that establish their decision-making patterns. Everything is tagged with provenance.

STEP 5

Interview your personas

Live interview: You ask questions in a real-time chat. The persona responds in character, drawing on their full memory, personality profile, and cognitive biases. A critic agent validates every response for consistency before delivery, catching contradictions instead of just generating plausible text.

Auto-interview: Configure an interview guide with sections and learning goals. Candor conducts the interview automatically, generating questions, tracking which learning goals have been covered, adapting probes based on responses, and detecting when a persona goes off-profile. It stops when coverage is achieved or gracefully moves on when a topic is exhausted.

STEP 6

Synthesis and reports

A seven-step synthesis pipeline processes every interview transcript. It extracts tagged signals (behavior, pain, attitude, constraint, goal, belief, preference, decision rule) and evaluates your assumptions against the evidence. Findings get clustered into themes ranked by frequency and intensity. The pipeline breaks down how different archetypes experienced the study and surfaces genuine tensions where personas disagree.

The synthesis output adapts to your study type. Problem discovery surfaces themes and tensions ranked by frequency and intensity. Problem validation returns explicit hypothesis verdicts (supported, contested, or not supported) with the supporting evidence. Concept testing frames what’s working versus what needs rethinking, by archetype. Price testing extracts value perception gaps, anchoring effects, and willingness-to-pay ranges. Layered value-prop and assumption validation return per-message and per-assumption verdicts on top of any of the above. Every claim links back to specific interview quotes.

What makes this different

Evidence grounding

Personas are built from your research documents and real market evidence, not generated from a prompt and imagination. Every attribute traces back to a source.

Full provenance

Every trait is tagged: grounded, inferred, calibrated, or low confidence. You always know why a persona holds a view, and you can trace any report finding back through themes to specific interview quotes.

Critic validation

A separate critic agent reviews every persona response before you see it. Hard contradictions are caught and regenerated. You get signal that stays in character, not an agreeable chatbot.

Common questions

Most studies run in 4 to 8 hours of elapsed time, with roughly an hour of your active attention spread across that window. Audience generation takes 25 to 35 minutes in the background (it's queued, you can close the tab). Persona generation runs in another 7 to 12 minutes once you approve segments. Interview guides generate in a couple of minutes. Live interviews happen in real time. Auto-interviews run in the background and finish based on persona count and interview depth. Synthesis runs automatically a few minutes after the last interview completes. The bottleneck is your time spent reading and reviewing, not Candor's processing time. End to end, you go from a research question to a structured report inside a single workday, instead of the weeks a panel study would take.

Your uploaded documents come first: customer interviews, survey data, market reports, or any research you already have. Candor supplements these with web evidence retrieval, searching for corroborating data about your target market. Every signal is tagged with its source. You always know whether a finding came from your data, public research, or inference.

No. Candor is built for UX research and insights leaders who want to scale their throughput, plus product teams, innovation teams, consultants, and founders who need customer signal without a dedicated research function. You don't need research training. Describe your audience in plain language, choose a study type, and Candor handles the methodology. Auto-interview manages question flow, coverage tracking, and signal extraction on its own. The synthesis report is structured to drive product decisions, not academic analysis.

Yes. Personas are persistent within a study. When you return to interview a persona again, they remember everything from prior sessions: specific stories they told, decisions they described, and how their views evolved. Show them a new concept and they'll react with full context of your conversation history.

Four primary interview types and two layered formats. Problem discovery explores unknown problems. Problem validation tests specific hypotheses against persona evidence. Concept testing evaluates what resonates and what needs rethinking. Price testing examines value perception, anchoring, and willingness-to-pay. Value-prop testing runs as a focused concept test or problem-validation interview to test which message lands. Assumption validation layers on any study type to return per-assumption verdicts. The synthesis pipeline adapts its output to match whichever type you choose. See the full set on the use cases page.

More FAQs →

Candor is in development.

Be the first to know when it launches.

No spam. Just a note when Candor is ready. Powered by Highline Beta.