How Architects Collect and Organize References
Architecture runs on precedent. Every project starts with reference gathering — how someone else handled a corner window, a brick soffit, a stair that doubles as seating — and the material accumulates across years and sources: ArchDaily and Dezeen tabs, Pinterest boards from three projects ago, site-visit photos on your phone, scans from monographs, screenshots of details you’ll want someday.
The result, in most practices, is fragmentation. The reference you know you saved is somewhere across four apps and two devices, and finding it costs more than re-googling it — which means your years of collected judgment quietly stop compounding.
Why the default setups fail
Pinterest boards are project-shaped, references aren’t. A board per project feels natural, but a great stair detail saved for the housing project is invisible when the school project needs a stair two years later. References outlive projects; boards bury them in one.
Folder trees force the wrong choice. Is that photo of a weathered
corten-and-larch facade filed under materials/corten, materials/timber,
facades, or precedents/pavilions? It’s all four. Whatever you pick, the
future search will come from a different direction — the full argument is in
tags vs folders, and
architectural references, which are always material + element + typology +
atmosphere at once, are a textbook case.
Site photos never leave the camera roll. The most valuable references you own — details you photographed yourself, buildings as they actually weather — sit untagged among lunch photos, on one device.
A workflow that compounds
1. One library, not one per project. Collect everything — web saves, site photos, scans — into a single pool. Projects then become views into the library rather than walls between its parts.
2. Tag on the axes you actually search. Three or four tags per image, drawn from a small fixed vocabulary:
- Element —
facades,stairs,openings,roofs,junctions - Material —
brick,clt,corten,concrete,render - Typology, when it matters —
housing,cultural,pavilions - Atmosphere, sparingly —
warm-light,monolithic,filigree
Now the corten facade answers corten, facades, and the pavilion search.
Add a project tag (proj-hillside) only when an image is actively in play.
3. Capture site visits into the same pool. The share-sheet habit: a detail catches your eye, photograph it, save it to the library with two tags before you’ve left the site. Thirty seconds, and the photo exists as a reference rather than camera-roll sediment.
4. Pull project sets when needed. Concept phase: search the axes the brief touches, tag the winners with the project tag, and you have a curated precedent set assembled from years of collecting — in minutes, not an evening of re-scrolling Pinterest.
Where Kelu fits
Kelu matches this workflow one-to-one: a single cloud library, capture from the iPhone share sheet on site, tags on as many axes as a reference needs, and search from the Mac at your desk or the web anywhere else. The same library on every device, first 2,500 items free — enough to move years of scattered references into one searchable place and see whether the system holds.
For the capture habit that keeps it alive, see why tags beat folders for screenshots; for calibrating your vocabulary, see how many tags should you use.