Research Database
A system for keeping what you read, so the next piece starts from accumulated material instead of a blank page.
Most people research the same things repeatedly because nothing from the last time survived.
The reading happened. The useful parts were noted somewhere — a document, a bookmark, a message to yourself. Six months later, writing on an adjacent topic, none of it is findable.
A research database is the fix, and the common failure is building one too elaborate to maintain.
Read the editorial process · Request Your Free Systems Audit
What it's for
Not storing everything you read. Storing what you concluded from what you read, in a form you can find later.
The distinction matters. A folder of saved articles is a reading list, and nobody rereads it. A database of your own observations, each linked to its source, compounds.
The unit: one note, one idea
The useful structure is small notes, each containing a single claim or observation, in your own words, with its source attached.
───────────────────────────────────────────
[A sentence stating the idea]
───────────────────────────────────────────
[Two to five sentences: what this means,
why it matters, what you think about it.]
Source: [title, author, link]
Date: [when you noted it]
Tags: [two or three]
Related: [links to other notes]
The title is the idea, not the topic. "Response time predicts conversion more than price" is findable and reusable. "Notes on sales article" is neither.
Your own words. A quoted passage you haven't processed is something you'll have to think about later. A note in your own words is thinking you've already done.
Capture and process separately
Two different activities, often conflated, and conflating them is why most systems collapse.
Capture happens while reading. Fast, messy, no structure. A link, a quote, a half-sentence. One place, zero friction.
Process happens separately — a scheduled slot, not whenever. Go through what you captured, turn the worthwhile items into proper notes, discard the rest.
Discarding most of it is correct. If you process everything you capture, you've built an archive.
Tagging that survives
Tags fail by proliferation. Keep two dimensions and nothing else:
Topic — five to ten broad areas, decided up front. Not thirty specific ones.
Type — what kind of thing this is: data, example, argument, counter-argument, quote, question.
The type dimension is underused and useful. When writing, you often need a specific kind of material — a concrete example, a number, an objection to address — more than a specific topic.
One rule: no new tag without deliberate thought. Tag sprawl is the most reliable way to make a research system unusable, and it happens gradually.
Linking
The single habit that turns a collection into something useful: when you write a note, link it to the related notes you already have.
This takes thirty seconds and does three things. It surfaces connections you wouldn't have noticed. It makes the database navigable without search. And it reveals where you have depth — a cluster of linked notes is an article that is most of the way written.
From notes to pieces
When you plan a piece, the first step isn't a blank page. It's gathering the relevant notes.
A well-maintained database means most pieces start with twenty to forty existing notes. The work becomes selection and arrangement rather than generation, which is both faster and better.
The signal that something is ready to write: you have a cluster of notes that argue toward the same point, and you can state that point in a sentence.
What to keep out
Things you haven't processed. An unread saved article is a to-do, not a note.
Everything you read. Selection is the value. A database with ten thousand entries has the same problem as the open web.
Client-confidential material. Keep that in the client's own space, with its own access rules.
Credentials, personal data, anything sensitive. The same rule as any knowledge system.
Maintenance
Less than you'd expect, if the structure stays simple.
Weekly: process the capture inbox. Thirty minutes.
Monthly: review the tag list, merge near-duplicates.
Quarterly: look for clusters — groups of linked notes that have become a piece worth writing.
That last one is the payoff. The database tells you what to write, based on what you've actually been thinking about.
Tool choice
Almost any tool that supports links between notes, search, and tags will work. The structure matters far more than the software.
Three things to check: can you capture in under five seconds, does search actually work, and can you link notes to each other.
Resist elaborate setups. Every feature you configure is something to maintain, and systems die from maintenance burden rather than missing capability.
Starting
Don't import your bookmarks. Don't process your archive.
Start with the next thing you read. Write three notes. Link them if they connect.
In three months you'll have a hundred notes you actually use, which beats a thousand imported ones you don't.
Store what you concluded, not what you read. The conclusions compound; the articles don't.
Want a content system that gets easier over time?
A free systems audit covers how your research and publishing actually work.