Built by researchers, for researchers
Heurista wasn't designed in a vacuum. It was built by an international development consultancy that has spent decades conducting evaluations in the field — and needed a better tool to do it.
I4DI — Washington, DC
I4DI (Innovations for Development and Impact) is an international development consultancy specializing in monitoring, evaluation, and learning. We've partnered with organizations including UNICEF, the World Bank, and USAID-funded programs across four continents.
Every methodology decision in Heurista — from the adaptive probing system to the bias control taxonomy to the evidence-tracing architecture — comes from real fieldwork. We built the tool we wished existed when conducting qualitative evaluations in resource-constrained environments with multilingual respondents and tight reporting deadlines.
Heurista is also used by Columbia University's Digital Learning and Teaching Services (DLTS) for educational research, demonstrating its applicability beyond international development into academic and institutional contexts.
Research-grade engineering
Heurista handles sensitive research data — informed consent records, personally identifiable information, and confidential interview transcripts. The engineering reflects that responsibility.
42 sequential database migrations
Foreign keys, indexes, constraints, and cascading deletes — evolved incrementally, not generated in one pass.
PostgreSQL with managed migrations
Production-grade Postgres with incremental schema migrations, foreign keys, indexes, and cascading deletes.
Row-level multi-tenancy
Organization isolation, team-based role access control, and study-level permissions enforced at the API layer.
29-point health monitoring
Automated subsystem checks across database, authentication, AI services, storage, and background processing.
Seven specialized AI agents
Each agent has its own methodology-trained system prompt — not generic AI instructions, but domain-specific guidance grounded in phenomenological bracketing, grounded theory constant comparison, emic/etic perspectives, and narrative temporal ordering. Model selection routes simple tasks to fast models and analytical reasoning to more capable ones.
AI Interviewer
Adaptive conversational interviews with methodology-specific probing across phenomenology, grounded theory, narrative inquiry, ethnographic, case study, and thematic analysis.
Dashboard Copilot
Data-driven chart selection with skewness detection, cardinality rules, Likert-scale handling, and audience-adapted layouts.
Hue Research Assistant
Conversational study assistant for data Q&A, cross-cutting pattern analysis, methodology guidance, and saturation tracking.
Qualitative Coder
Three-phase pipeline: AI proposes codes, QA review catches errors, cross-session synthesis surfaces themes.
Survey Designer
Generates complete research instruments from descriptions with question types, conditional logic, and section structure.
Report Generator
Multi-pass synthesis: per-theme analysis, per-question analysis, and cross-cutting findings with evidence tracing.
Question Generator
Creates professional survey instruments from research objectives with methodology-appropriate interviewer guidance.
Six-point bias control taxonomy
Every AI-conducted interview is governed by six distinct bias controls. The adaptive probing system also detects five types of response quality issues — surface-level answers, evasive responses, social desirability, contradictions, and emotional distress — and adjusts technique for each.
Leading questions
Never suggests answer directions
Social desirability
Probes performative answers for concrete examples
Acquiescence
Varies framing to avoid yes-bias
Anchoring
Never references other respondents or expected ranges
Confirmation
Equal follow-up for surprising and expected answers
Interviewer neutrality
Consistently neutral tone across all responses
Security built for sensitive data
Research data demands more than default security. Heurista uses iron-session with AES-256 encrypted cookies, scrypt password hashing with unique salts, CSRF protection, and timing-safe token comparison. Automated encrypted backups run daily. AI processing follows data minimization principles — only the context required for each task is sent to the configured AI provider, and your research data is never used for model training.
For organizations requiring full data sovereignty, Heurista offers enterprise private cloud deployment where AI processing runs entirely within your own AWS or Google Cloud infrastructure — no research data leaves your environment.
GDPR and CCPA compliant with a published Data Processing Agreement, privacy policy, and AI transparency disclosure. Full audit trails track every action.
See it for yourself
14-day free trial. Demo studies included so you can explore the full platform immediately.
Start researching free