Research that lives only in the researcher's head is a single point of failure. It walks out when they do, and in the meantime it loses arguments to whoever is most confident in the room, because nobody else can check it.
Two jobs. Synthesis compresses hours of raw material into a handful of findings specific enough to design against. A repository keeps those findings — and the evidence under them — retrievable by someone who was not in the sessions.
Neither is admin. Skip them and you pay later: the same interviews run for the third time, and a team deciding what users want from memory.
Synthesizing Research
Work from the raw material, not your memory of it. Break transcripts, session notes and open-text survey answers into single observations — one per card, whatever your tool is — each tagged with its participant, so any claim traces back to a person.
Then group them, and let the group names come out of the material rather than the buckets you arrived with.
Agree a threshold before you start. One person in eight is a quote; five in eight is a pattern. Where you set the line between them depends on what rests on the finding — a reversible decision can run on two, a rebuild cannot. Write each finding as a claim with its evidence underneath.
Synthesise with at least one other person — ideally two: one who ran the sessions, one who did not. Done alone, this step hands you back the findings you walked in with.
- Synthesizing UX Research: Making What's "Mysterious" Clear — article
- How to synthesise information — article
Research Repository
One searchable place holding clips, transcripts, survey data and written findings. The test is concrete: a new joiner asks what we know about onboarding for small teams and has an answer in five minutes without messaging anyone.
That needs tagging, and tagging needs a short controlled vocabulary someone owns — segment, product area, method, date, study. Let anyone invent a tag and you get five words for the same thing, and a search that quietly stops returning what is already there.
Store the evidence beside the conclusion — the clip, the transcript line, the survey rows. Without it a finding can only be taken on trust or researched again when the product moves underneath it. Put a date on every study and treat old ones as describing a product that no longer exists.
An unmaintained repository is worse than none, because people still trust it.