How to Build a Tagging System for Screenshots

This is the hands-on guide. If you want the why first — why tags and not folders, why fewer tags beat more — those arguments live in tags vs folders and why tags beat folders for screenshots. Here we just build the system, in four steps you can finish this week.

Step 1: Write your starter vocabulary (15 minutes)

Don’t tag anything yet. First, decide the words — because a tag system dies by improvisation: mint tags on the fly and in six months ui, UIs, and interface are splitting your search three ways.

Open a note and write three short lists, drawn from what you actually save:

  • Patterns — the kinds of things: pricing, onboarding, empty-states, settings, emails, landing-pages
  • Qualities — why you save them: playful, minimal, data-dense, dark-mode, copywriting
  • Sources / projects — only ones you’d filter by: linear, client-x

Aim for 15–25 tags total. The test for each: can you picture ten future screenshots wearing it? If not, cut it. And fix the conventions now — lowercase, single words, plural — so empty-states never grows an Empty State twin. (Full sizing logic: how many tags should you use.)

Step 2: Set up the capture habit (one day)

The system runs on two separate moments, and keeping them separate is the whole trick:

  • Capture is instant and thoughtless. Screenshot → share sheet → library. No tagging, no deciding. Under five seconds, or you’ll stop doing it.
  • Tagging is batched. Once a week, ten minutes: open the untagged inbox, give each screenshot 2–4 tags from your vocabulary, delete the ones that no longer spark anything.

Trying to tag at capture time is how systems die — it taxes the exact moment you have the least attention. Batching moves the cost to a moment you chose.

Step 3: Tag with the 2–4 rule

Per screenshot: one pattern tag, one or two quality tags, a source tag if it earns it. That’s it. Two prompts keep you honest during the weekly pass:

  • “What would I type when looking for this?” Tag that. Not what the image contains — what future-you will ask for.
  • “Does this tag distinguish it?” If a tag applies to half your library, it’s dead weight. Skip it.

Resist retro-tagging your entire backlog in one heroic session. Tag the new stuff weekly; tag old items lazily, whenever a search surfaces them untagged. The backlog matters less than the habit.

Step 4: Prune quarterly (10 minutes)

Four times a year, scan your tag list:

  • Merge synonyms — two tags meaning one thing: keep one, migrate, delete.
  • Kill orphans — a tag with two items after three months was a bad mint.
  • Split the overgrown — one tag holding 40% of the library isn’t narrowing anything; split it along the line you actually search (ui → onboarding / settings / pricing).

This is the maintenance budget: forty minutes a year. A folder hierarchy costs more than that per week.

Where Kelu fits

You can run this system anywhere that has real tags and real search, but Kelu is shaped exactly for it: share-sheet capture from the iPhone in seconds, an untagged inbox for the weekly pass, as many tags per screenshot as it needs, and search across the same library from iPhone, Mac, and the web. First 2,500 items free — more than enough runway to find out that the ten-second retrieval test (“that pricing page with the toggle” → on screen) actually holds.