Research Synthesis

Transform raw research evidence into a traceable synthesis without turning model inference into user fact.

Working rules

• Preserve provenance. Give each source a stable label and keep insights connected to quotes, observations, or records.
• Keep four layers distinct: raw evidence, pattern, interpretation, and recommendation.
• Do not invent participants, quotes, counts, personas, motivations, or consensus.
• Report frequency only within the supplied sample. Do not generalize a small or biased sample to the whole market.
• Retain contradictory evidence, minority behaviors, and negative cases; they often reveal segmentation or risk.
• Merge semantically duplicate feedback while preserving differences in context, severity, and user type.
• Redact or minimize unnecessary personal and sensitive information in the output.
• Do not publish findings or update external repositories unless explicitly asked.

Synthesis workflow

1. Inventory the evidence: source labels, research method, participant or channel, relevant segment, date, and known sampling limitations.
2. Extract atomic observations. Each item should express one behavior, need, pain point, workaround, trigger, or outcome and retain its source label.
3. Normalize and cluster observations by meaning rather than shared vocabulary. Avoid forcing every item into a theme.
4. For each theme, describe:
   • the observed pattern;
   • supporting and contradicting evidence;
   • affected users or contexts;
   • strength and limitations of the evidence;
   • the interpretation, clearly labeled as interpretation.
5. Derive insights only when they explain why the pattern matters for user behavior or outcomes. Avoid merely restating observations.
6. Translate supported insights into opportunity areas or testable hypotheses. Do not jump directly to features unless the user asks for ideation.
7. Identify research gaps and the cheapest next study or question that would reduce the most consequential uncertainty.

Evidence strength

Use qualitative labels rather than false precision:

• Strong within sample: repeated across multiple independent sources with little contradiction.
• Moderate: repeated but concentrated in a segment or accompanied by meaningful counterevidence.
• Tentative: based on one or few sources, indirect evidence, or an inference requiring validation.

Explain the basis whenever a label could affect prioritization.

Default deliverable

Adapt to the volume of material, normally using:

1. Research scope and sample limitations
2. Executive synthesis
3. Evidence map: source, observation, context, and theme
4. Themes and insights with supporting and contradicting evidence
5. Segments, tensions, and outliers
6. Opportunity areas or hypotheses
7. Open questions and recommended next research

If the source material is too thin for a defensible insight, provide an evidence inventory and research gaps instead of manufacturing conclusions.