Turn research fragments into useful notes

The retrieval test

Before changing anything, try a simple experiment. Pick a topic you read about three months ago and search your notes for it. Time how long it takes to find the key source and the point you took from it. If you cannot find it within a minute or two, your notes are collecting well but retrieving badly. That is the problem this page addresses.

Retrieval fails for a handful of predictable reasons. The note has no title, or a title only you-on-that-day could understand. The source is missing, so even when you find the note you cannot cite or re-read it. The note mixes several ideas, so it matches too many searches or none. Or it sits in one of several inboxes you have forgotten exist.

Three kinds of note, in plain terms

Different writers use different names, but most working systems separate three kinds of material:

Source notes record what someone else said, with enough detail to find it again: author, title, year, page or timestamp, and a short quotation or summary. Sönke Ahrens, in How to Take Smart Notes (2017), calls these literature notes.

Working notes are your own thinking in progress: questions, tentative comparisons, drafts of an argument. They are allowed to be messy and dated.

Settled notes state one idea in your own words, clearly enough that you could understand it in a year. Andy Matuschak describes a similar idea as evergreen notes. These are the notes worth linking carefully.

You do not need all three on day one. Most people benefit first from simply separating what a source said from what they think about it.

One landing place, emptied on a schedule

Pick a single capture inbox. It can be a notes app folder, a text file or a paper notebook page. Everything goes there first: web highlights, voice memo transcripts, photos of whiteboards. The rule is not that the inbox stays empty but that you look at it on a known day. Weekly is enough for most people. When you process an item you either give it a title and source and file it, or delete it. Deleting is a legitimate outcome; not every fragment deserves to survive.

Before and after: an annotated example

The note below is fictional and clearly labelled. It shows how a typical captured fragment changes when it is processed.

meeting w/ Sam — he said the pilot didn't work bc onboarding?? check later

Title: Pilot drop-off linked to onboarding length (fictional)
Date: 2026-02-03
Source: Project call, notes by me
Claim: Users left during a 40-minute setup
Status: unverified — ask for usage data
Related: [[Onboarding length and retention]]

Titles do most of the work

Search engines inside note apps weight titles heavily, and so does your eye when scanning a list. A good title is a short claim or question: "Spaced review helps vocabulary more than grammar" rather than "Spacing". If a note contains two claims, it probably wants to be two notes. This is the practical heart of the atomic-note principle associated with Niklas Luhmann's Zettelkasten, a slip-box of around 90,000 cards he used over several decades of sociological writing.

Paper is welcome

Handwriting helps many people think. The cost is that paper cannot be searched. A light fix is to keep a running digital index: one line per notebook page with the date, page number and a descriptive title. Methods such as the Cornell note-taking layout, described by Walter Pauk in How to Study in College, already reserve a cue column that converts neatly into index lines.

Realistic limits

Processing notes takes time, and you will fall behind. That is normal. If your inbox grows beyond what you can face, archive everything older than a chosen date into one folder, keep it searchable, and process only new material. You lose nothing, and you regain a system you are willing to open.