Building a research repository people actually open

Every research team eventually builds a repository, and most repositories eventually become a place research goes to be safe from readers.

The usual post-mortem blames tagging discipline, or the tool, or “adoption.” But teams with immaculate taxonomies still end up with a beautiful archive nobody opens. The problem is usually one level down: repositories store documents, and the people you want reading research don’t want documents. They want answers.

Store claims, not files

A deck titled Q2 Onboarding Study answers no question anyone is currently asking. A claim — “New admins abandon setup at the permissions step because they’re afraid of breaking things” — answers one that comes up in every planning cycle.

The unit of a living repository is the claim, backed by evidence: the quote, the clip, the moment in the transcript where the customer says the thing. Files are where claims live; they shouldn’t be the interface.

Make the evidence one click deep

Trust in research is built by proximity to the source. When a PM can click from “customers fear the permissions step” to four customers saying exactly that, two things happen: they believe it, and next quarter they come back without being invited.

Every layer between a claim and its source — “see slide 14”, “it’s in the June folder” — costs you a reader.

Let questions be the front door

Nobody browses a repository for fun. They arrive with a question: have we heard anything about export limits? What do enterprise admins think of SSO setup? If the front page is a folder tree, they leave. If it’s a search box that answers questions with evidence, they stay.

This changes what “organizing research” means. You’re not filing documents; you’re pre-answering the questions your team asks every quarter.

Retire your findings

Evidence ages. A repository that treats a 2023 quote and a last-week quote as equally alive will eventually mislead someone, get caught, and lose the team’s trust in one afternoon. Date everything, surface recency, and archive claims when the product has moved beneath them.

---

None of this requires a bigger team. It requires storing the right thing. Capture interviews so the evidence is quotable, let claims carry links to their sources, and put a search box where the folder tree used to be.

A repository earns its existence the first time someone you didn’t invite quotes it in a meeting. Build for that moment.