How to Run User Interview Analysis in Dovetail
Conducting user interviews is only half the research process. The other half, analysis, is where insights emerge. Most teams skip rigorous analysis because it is time-consuming and the tools are clunky. They end up with bullet-point summaries in a Google Doc that reflect the interviewer's biases more than the data. Dovetail is purpose-built for research analysis: it gives you the tools to tag transcripts, cluster patterns, and surface insights with evidence that the team can verify.
This guide walks through the end-to-end analysis workflow in Dovetail, from importing interview transcripts to publishing insights that influence your product roadmap. The key principle is that every insight should be traceable back to specific moments in specific interviews, so when a stakeholder challenges a finding, you can show them the evidence rather than asking them to trust your summary.
Step-by-step guide
Set up your Dovetail project and import transcripts
Create a new project in Dovetail named after your research initiative (e.g., 'Q3 2026 Onboarding Research'). Import your interview transcripts either as text (copy-paste from your transcription tool) or by uploading video/audio files directly for Dovetail to transcribe. Add each participant as a separate note with metadata: participant name or ID, date, persona segment, and any relevant demographic or behavioral data. Group notes by interview type or participant segment.
- Create a project with a descriptive name tied to the research initiative
- Import transcripts or audio/video files for each interview
- Add participant metadata: segment, date, and relevant context
Create your tag taxonomy
Before coding transcripts, define your tags. Create a two-level taxonomy: top-level themes (e.g., 'Onboarding Pain,' 'Feature Discovery,' 'Pricing Perception') and sub-tags within each theme. Add a sentiment dimension: Positive, Negative, Neutral. Keep the initial taxonomy to 5-8 top-level themes. You will add more as patterns emerge, but starting too granular makes coding slow and inconsistent. Share the taxonomy with anyone who will help code transcripts so tagging is consistent across coders.
- Define 5-8 top-level theme tags based on your research questions
- Add sub-tags for expected subtopics within each theme
- Add sentiment tags: Positive, Negative, Neutral
- Document tag definitions so multiple coders tag consistently
Code your transcripts by highlighting and tagging
Read through each transcript and highlight relevant passages. For each highlight, apply one or more tags from your taxonomy. A single passage can have multiple tags if it touches multiple themes. As you code, you will notice themes that do not fit your existing taxonomy. Add new tags as needed, but keep the taxonomy tidy by merging similar tags periodically. Aim to code at least 3-4 interviews before looking for patterns so you are not anchored by the first interview.
- Highlight key passages in each transcript
- Apply relevant tags to each highlight
- Add new tags as unexpected themes emerge
- Code 3-4 interviews before looking for patterns
Identify patterns using Dovetail's analysis views
Use Dovetail's tag board to see all highlights grouped by tag. Look for patterns: which tags have the most highlights, which themes appear across multiple participants, and which tags have strong sentiment skew. Use the participant matrix view to see which themes each participant mentioned. A pattern is reliable when it appears across 3+ participants independently. Single-participant themes are anecdotes, not patterns. Document the patterns you find with a count of supporting evidence.
- Review the tag board for high-frequency themes
- Use the participant matrix to verify patterns appear across participants
- Distinguish between patterns (3+ participants) and anecdotes (1-2)
Write insights with evidence chains
For each pattern, create an Insight in Dovetail. An insight has three parts: the finding (a declarative statement like 'Users abandon onboarding when asked to connect integrations before experiencing core value'), the evidence (links to the specific highlights from 3+ interviews that support it), and the recommendation (a suggested product action). The evidence chain is what makes research trustworthy. Without it, insights are opinions with a research label.
- Write insights as declarative statements, not questions
- Link each insight to the specific tagged highlights that support it
- Add a recommendation for the product team
Publish and share insights with stakeholders
Use Dovetail's sharing features to publish insights to your team. Create a highlight reel of the most impactful video clips (if you recorded interviews) for stakeholders who will not read the full analysis. Share the project link in your product channel and reference specific insights in your next sprint planning or roadmap review. The goal is to make research findings easy to discover and cite. Every time someone references 'that research about onboarding,' they should be able to link to the specific Dovetail insight.
Common mistakes
Summarizing instead of coding
Writing a paragraph summary of each interview is faster than tagging individual passages, but it loses the evidence chain. Summaries reflect the summarizer's interpretation; tagged highlights preserve the participant's actual words. When the team debates whether a finding is real, you need the raw data, not a summary.
Confirming existing assumptions rather than exploring the data
If your tag taxonomy only contains themes you expected to find, you will only find what you expected. Leave room for unexpected themes by coding with an open taxonomy and adding new tags when participants say something surprising. The most valuable insights are often the ones you did not anticipate.
Drawing conclusions from too few interviews
Five interviews is the minimum for identifying patterns. Drawing product conclusions from 2-3 interviews is unreliable. If you only have 2-3 interviews, label your findings as 'preliminary signals' rather than 'insights' and plan follow-up research to validate them.
Tips
Use Dovetail's AI transcription feature to save time on manual transcription, but always review the transcript against the audio for accuracy on domain-specific terminology
Create a 'Research Repository' workspace in Dovetail where all completed projects live, making past research discoverable for future initiatives
Tag insights with the product area they affect so PMs working on different parts of the product can filter for relevant findings
Schedule a 'Research Readout' session within one week of completing analysis while the context is fresh and stakeholders are still interested
How Vantage helps
Vantage can ingest your research insights as context for PRD generation. When you link your research findings, Vantage weaves them directly into the requirements and user stories, ensuring that specifications are grounded in evidence from real user conversations rather than assumptions. The research stops being a deliverable and becomes an active input to every product decision.