UX Research · Step 7 · 6 min read
Research repositories
A research repository is the team's memory: one place where studies, insights and evidence can be found again. What goes in one, the three ways to build it, what atomic research and nuggets are, and why most repositories fail, with the numbers from a survey of more than 400 practitioners.
A study nobody can find again is a study you'll pay for twice. Six months after a project, someone asks "what do we know about how people split bills?". If the answer lives in one researcher's memory and a deck in their personal drive, the team either asks again or guesses. A repository is how research stops evaporating.
What you'll be able to do after reading:
- Say what a research repository is for, and what goes in it
- Choose between a document library, an insight database and a hybrid
- Write an atomic research nugget with its evidence and tags
- Avoid the four reasons repositories fail
1. What it is, and what goes in
A research repository is a central place to store research outputs so others in the organisation can reach them. It helps people find insights faster, follow a topic across studies over time, and avoid running the same study twice (Maria Rosala, NN/g, 2024).
What goes in varies. NN/g lists reports, standalone insights, study materials such as plans and screeners, recordings, clips and transcripts, and raw notes, and says plainly that "there's no hard rule on what belongs in a research repository". In its survey of 411 practitioners, reports were the most stored item, in about 70% of repositories (NN/g, 2024).
One rule that isn't optional: recordings and transcripts are personal data. Store them with access limits and a deletion date that matches what participants agreed to. See Research data and privacy.
2. Three ways to build one
Whole reports, plans and recordings, in folders by product or year.
Easy to add to. Hard to find one insight in.Separate insights, each tagged, searchable on their own.
Easy to find. Every entry costs effort, and can lose its context.Tagged insights, each linking back to the study, the plan and the evidence.
What most lasting repositories become.The trade-off is between how easy it is to find a single insight and how much work it is to put one in (NN/g, 2024). A folder of reports is easy to add to and hard to search. A database of separate insights is easy to search and costs effort every time someone contributes. Most teams that last end up with a hybrid: each insight links back to the study it came from.
NN/g's survey found teams using collaboration tools such as Confluence and SharePoint (43%), research platforms such as Dovetail or Condens (32%) and database tools such as Notion and Airtable (14%) (NN/g, 2024). The tool matters less than the structure, but it does matter: see section 4.
3. Atomic research and the nugget
People check a split in the group chat before they trust it in the app
Clip, 0:34 · P4 scrolls the flat's chat to find the photo of the electricity bill, then redoes the sum on her phone's calculator · Interview 4, 23:15–23:49
Tomer Sharon, then running research at WeWork, argued in 2016 that "a report is not the atomic unit of a research insight". His unit was the nugget: one observation from research, tagged so it can be found later (Sharon, The atomic unit of a research insight, 2016). In 2018 he described atomic research as "a tagged observation supported by evidence", with the evidence ideally a short video snippet of 15 to 45 seconds, and tags that can be procedural (date, method, source), about the experience (magnitude, frequency, emotion) or about the business (product line, journey) (Sharon, Foundations of atomic research, 2018).
Why it's useful:
- Insights outlive their study. A nugget from a churn interview can answer a pricing question a year later.
- Evidence travels with the claim. The timestamp lets a sceptic watch the 30 seconds themselves.
- Patterns appear across studies. Ten nuggets with the same tag from four studies is a stronger case than one report.
And the cost: someone has to write and tag every nugget, consistently. Sharon himself noted that at WeWork "nuggetizing" took about as long as writing a traditional report, and paid back later by answering questions without new studies (Sharon, 2018). Start with a small, fixed set of tags and grow it only when people ask for one.
4. Why repositories fail
NN/g surveyed more than 400 UX and ResearchOps professionals about their repositories. Only 9% described theirs as mature and thriving, 11% said stagnant or forgotten, and fewer than half rated adoption as good. Ownership was the clearest gap: 29% of repositories had no owner at all (Maria Rosala, NN/g, 2024).
The four reasons the article gives, and what to do about each:
| Reason | What to do |
|---|---|
| Very low UX maturity | Start with what the team already asks for: one topic, one product |
| Frustrating tools | Pick a tool people already open every day, or one built for research; test it with the people who'll search it |
| No owner | Name one person, with time in their week for it |
| Too much work to contribute | A template, a short tag list, and contributing as the last step of every study |
It also takes time: good adoption was reported by 65% of repositories older than three years, and about a third of the younger ones (NN/g, 2024). Treat the repository as a product: research what its users search for, test its structure with a card sort or tree test, and keep improving it (NN/g, 2024).
5. Common mistakes
- Buying a tool before deciding what the repository is for.
- Uploading every old report on day one, untagged.
- Fifty tags, each used once.
- Insights with no link back to the study and the evidence.
- No deletion dates for recordings.
- Nobody owning it, so it quietly dies.
6. Vocabulary
| Term | Meaning |
|---|---|
| Research repository | A central, searchable store of research outputs and evidence |
| Document library | A repository of whole reports and files, usually in folders |
| Insight database | A repository of separate, tagged insights |
| Atomic research | Storing research as small, tagged observations with evidence |
| Nugget | One observation from research, with its evidence and tags |
| Taxonomy | The agreed set of tags and categories used to organise the repository |
7. Try it (no tools needed)
- Take one study you ran or read.
- Write three nuggets from it: one observation each, the evidence (quote and timestamp), and three tags.
- Ask a colleague to find "anything about trust" using only your tags.
Reflection questions: Which tags would someone outside your team actually search for? Who in your organisation would own the repository, and when would they do it?
8. If you already practise
- Audit your repository's search: give three people three real questions and watch them look.
- Add a "last checked" date to insights, and retire the ones the product has outgrown.
- Make contributing part of the definition of done for every study.
Checklist
Learn more
- Maria Rosala, Research Repositories for Tracking UX Research and Growing Your ResearchOps (NN/g, 2024)
- Maria Rosala, Why Research Repositories Fail and How to Get Them Right (NN/g, 2024)
- Tomer Sharon, The atomic unit of a research insight (2016) · Foundations of atomic research (2018)
Books, courses and videos to go deeper
- Research That Scales: The Research Operations HandbookKate TowseyBook · Rosenfeld Media, 2024 (Amazon)
- It's Our ResearchTomer SharonBook · Morgan Kaufmann, 2012 (Amazon)
Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.
Related
Research reports people read · ResearchOps · Research data and privacy · Thematic analysis