The synthesis process is where research either pays off or falls apart. Raw interview notes — even good ones — are not insights. They are the raw material from which insights are extracted through a structured analytical process. This walkthrough covers each step from first transcript to deliverable report, with decision points and common mistakes at each stage.
Step 1: Prepare Your Materials
Before analysis begins, assemble all transcripts or notes in a consistent format. Number each participant with a code (P1, P2, etc.) rather than a name — this protects participant privacy in shared documents and prevents the halo effect of attributing extra weight to participants whose names you remember. Note each participant's segment, role, and relevant context at the top of their transcript.
If you are working from audio or video, generate transcripts before coding. Coding from memory or from listening is slower and introduces recall bias. Written transcripts are the only reliable input to systematic thematic analysis.
Step 2: First-Pass Reading
Read all transcripts once without coding. The purpose of this pass is orientation — understanding the overall shape of the dataset before committing to a coding structure. Note observations that surprise you, recurring phrases, and anything that seems inconsistent with your prior assumptions. Do not create codes yet.
This step is frequently skipped under time pressure. Skipping it increases the probability of anchoring on early transcripts and missing the pattern that only becomes visible after reading the full dataset.
Step 3: Open Coding
Read through the transcripts again, this time applying descriptive codes to meaningful passages. At this stage, codes should be close to the participant's own language — "can't find the export button" rather than "navigation friction." Keep codes granular. You will group them later; you cannot un-group overly broad initial codes.
Mark each coded passage with the participant code and a timestamp or page reference so you can return to the original context when needed.
Step 4: Theme Development
Group related codes into candidate themes. Look for patterns across participants — a theme is something that appears in multiple transcripts, not a single striking observation from one participant. Give each theme a clear name that captures the underlying experience, not just the surface topic.
Challenge every theme: does this represent a meaningful pattern, or am I forcing codes together because they superficially sound similar? A theme about "onboarding confusion" and a theme about "feature discovery difficulty" might be the same problem or two different ones — the distinction matters for recommendations.
Step 5: Writing the Report
Structure the report around findings and recommendations, not around methodology. The methodology section belongs in an appendix for readers who need to evaluate the rigor of the research. The main body should answer the research questions with clear findings, each supported by two to three illustrative quotes and a specific recommended action.
A finding without a recommendation is incomplete. A recommendation without an evidence-backed finding is an opinion. The combination is what makes research useful to the people who will act on it.