Research Synthesis: From Notes to Actionable Insights

What is it, and what is it for?

After five interviews, a team usually has dozens of notes, quotes, and screenshots. Without synthesis, each person remembers whatever impressed them most, and the team argues opinions instead of evidence. Synthesis forces you to look at everything together and ask what keeps repeating.

The most common method is the affinity diagram. NN/g describes it in three moments: generate notes, organize them into clusters, and prioritize the clusters and next steps (NN/g: Affinity Diagram). This page covers the whole path, from notes to insight. For the team workshop format, see Affinity Mapping.

What is the step-by-step process?

The guides behind this page (NN/g, User Interviews, and LoopPanel) agree on a similar sequence:

1. Turn each observation into a note. User Interviews asks you to break notes, transcripts, and feedback into single observations on movable notes, physical or digital, and highlight interesting phrases. If you cut the data too much at this step, you have less material and more analysis to do later.
2. Group by similarity. Start with one note, add related notes, or open a new cluster. You can begin with themes tied to your research questions and name each cluster as you go.
3. Name the clusters. LoopPanel recommends concise, legible labels. In its e-commerce example, clusters get names such as “Navigation,” and the checkout process shows up as a key area to improve.
4. Repeat, and restart if needed. Group again so different patterns can surface. User Interviews suggests photographing or copying earlier rounds so nothing is lost. LoopPanel asks you to be willing to regroup until the arrangement holds together.
5. Split by context. Sort responses by participant attributes such as role, age, or location, if your study captured them. According to User Interviews, that data can be key to discovering hidden patterns.
6. Zoom out and prioritize. Look at how the clusters relate and which ones matter more. NN/g proposes reviewing each cluster, documenting next steps, voting on the most important ones (for example with dots), and naming a point person for the actions.
7. Turn clusters into insights. Each cluster becomes a short summary of what you learned.

How do you write an insight?

An insight is not a list of complaints. According to NN/g, some teams use different note colors to separate what the person said (direct quotes) from what the researcher saw (observations). That makes it clear what is data and what is interpretation. User Interviews suggests turning cluster names into short summaries to share with the team, and converting findings into outputs such as a “Jobs to Be Done” list, a prioritized list of pain points, or a user journey map.

Illustrative example of the difference between an observation and an insight:

• Observation: “Five of six participants looked for the price on the home page before entering the catalog.”
• Insight: “Price is a question people ask before deciding to buy, not a detail that belongs on the product page.”

The second statement helps you decide. The first only describes.

What biases can synthesis hide?

User Interviews notes that affinity mapping helps counter confirmation bias (favoring the data you already expected) and recency bias (favoring the most recently collected data), because it makes you look at all the notes. But the method does not remove them. LoopPanel warns about two traps: prejudging the data and imposing your own interpretation, and getting attached to your first groupings or labels. NN/g adds others: dominant voices in the session, forcing notes into a cluster, dismissing small clusters, and fatigue if you arrive at the workshop with too much untrimmed data.

Before you trust a pattern, review research bias and the sample size in sample size.

When should you NOT run a full synthesis?

• If the study has one session, the full process is excessive. It is enough to write down what you observed and compare it with the study’s objective.
• If your question is about how many or how much, an affinity diagram cannot answer it. You need quantitative data, and a survey such as NPS, CSAT, and CES may be the complement.
• If the team will not use the synthesis to decide anything, the work stays on a board. Before you start, define which decision it must support.

Illustrative example

Illustrative example: a team interviews eight people who canceled a subscription. When the notes go onto the board, three large groups appear: “I don’t understand what the plan includes,” “I was charged without warning,” and “I wanted to pause, not cancel.” When the team regroups, a fourth cluster emerges: “I looked for a pause option and could not find it.” The insight is that pausing exists, but people do not perceive it as an alternative to canceling at the moment they decide. The team does not redesign the plan; it moves the pause option to the cancellation screen and tests again with five people.

If your synthesis feeds a design, you may later need proto-personas to summarize who each insight applies to. The research plan should say in advance which questions you are trying to answer, so you do not end up with a board full of interesting things that do not help. And to take the results to other people, see presenting research findings and research repository.

Common mistakes

• Synthesizing alone and presenting conclusions the team has not discussed.
• Writing observations and calling them insights.
• Grouping by keywords instead of by the problem the person is living through.
• Forcing a number of clusters: there is no correct number.
• Mixing user interviews with stakeholder interviews without keeping them apart.
• Leaving the synthesis on the board instead of turning it into a decision or a test.

What you can do tomorrow

1. Move the notes from your last session onto digital sticky notes, one observation per note.
2. Group them twice: the second time from scratch, and keep a screenshot of the first version to compare.
3. Write three insights in the form “observation, implication” and ask the team which one would change a decision.

Sources

• NN/g: Affinity Diagram
• User Interviews: Affinity mapping for UX research data synthesis
• LoopPanel: Affinity mapping guide