[ In progress ]
built-asset information management system
I designed the naming canon, ETL and governance rules that are turning a decades-old archive into an auditable base. Public sector, operational phase, ISO 19650-3.
- 1/3
- of the archive was redundant copies (MD5)
- 40
- dated snapshots over 89 days
- −14%
- fewer items after cleanup, so far
- 0
- files deleted: everything moved, nothing lost
context
A public-sector body that keeps the register of built assets: dozens of buildings across several campuses, and decades of documentation piled into a shared drive with no common grammar.
problem
Nobody could know with certainty, at any moment, what existed without asking a specific person. Knowledge lived in heads, not in the system, and every absence or change of staff opened a gap no one could measure.
decisions
-
Grammar before movement
No file moves until the naming convention exists in writing. Tidying without a grammar first only moves the mess somewhere else.
-
Space as the axis, topic as the shortcut
The atom is the built unit and the organisation follows real geography. Thematic views exist as links, never as copies.
-
Every action reversible and provable
Immutable snapshots, with MD5 as proof. Nothing gets deleted by accident, only moved, and the diff between two states is the record.
-
Automation proposes, a person executes
Scripts write the proposal and the log; anything destructive goes through a human hand.
system
A closed loop: photograph the state, compare it to the previous one, decide on the difference, execute through a reviewed script, log the dated decision. Once a month, the harvest turns the log into documents: dossier, status, handover.
ID _ BUILDING _ DATE _ TYPE _ VERSION
outcome
Knowing what exists, and where, no longer depends on asking someone. The system keeps its own record.