Methodology essays, opinion pieces, and lessons from building research-grade synthetic personas. New entries publish on a regular cadence.
2026-07-07 • 9 min read
Synthetic-then-human is one way to use synthetic research, not the only one. Three patterns teams actually run, and how to tell which one fits your question.
2026-06-30 • 9 min read
The strongest case against synthetic users is worth taking seriously. Most of it argues for grounding and discipline, not for skipping synthetic research.
2026-06-23 • 11 min read
Most AI persona tools have no memory. Each session resets. Candor's persona memory architecture (six memory types grounded in cognitive science) makes a synthetic persona behave as the same person across every interview session within a study.
2026-06-16 • 14 min read
Most "best of" lists in this category are bad. Here's a framework for evaluating synthetic research platforms honestly, plus the major categories of tool you'll encounter in 2026, named and characterized without ranking.
2026-06-11 • 10 min read
Most products ship on assumptions nobody pressure-tested. Surface them, validate them with synthetic research, rewrite the spec. A framework for product teams that want fewer features built on quietly wrong premises.
2026-06-09 • 11 min read
Two synthetic personas with the same demographics can still reason completely differently. The difference is personality. Here's how Candor uses OCEAN, calibrated by region and occupation, to shape research-grade reasoning.
2026-06-04 • 11 min read
Big brands test 10 to 50 concepts per project at $15K to $50K each. Most get killed in a brand-team meeting before any respondent sees them. Synthetic-first sequencing fixes the math.
2026-06-02 • 10 min read
Synthetic concept testing is one of the strongest use cases for synthetic research and one of the easiest to misuse. Here's the decision framework, plus the four conditions where it earns its keep and the four where it falls apart.
2026-05-29 • 16 min read
Real audiences aren't one homogeneous group. Archetype clustering surfaces the distinct types of person inside them from real evidence: 3 to 8 clusters per study, each with calibrated personality and bias ranges.
2026-05-27 • 14 min read
Evidence grounding is what makes a synthetic research output you can audit and defend in a real decision. Here is how the pipeline actually works.
2026-05-21 • 6 min read
Most synthetic user tools generate personas from a description and call it research. Evidence grounding is the difference between a research instrument and a confident guess.
Be the first to know when it launches.
No spam. Just a note when Candor is ready. Powered by Highline Beta.