Provenance: keeping the story of the data
Track where a record came from and what happened to it so another person can understand, trust, or reproduce the result.
Why this matters
Two records with the same value can carry very different credibility if one came directly from a calibrated sensor and another was copied through several transformations. Without provenance, that history disappears.
Learn it
Provenance can include source system, source record ID, observation time, ingest time, device or human source, method, transformation history, software or workflow identity, and references to the original data. The exact fields depend on the project, but the principle is stable: keep enough lineage to explain the current record.
Work through an example
If a KML point is converted to GeoJSON and then normalized into a workspace dataset, provenance can record the original file, conversion step, import time, and source identifier.
source β transformation β normalized record β derived output
Keep IDs and timestamps at each important step.Practice
- 1Trace one record backward from its current dataset to its original source.
- 2Write down each transformation it passed through.
- 3Ask whether another person could reproduce or audit the result from your notes.
Try it
Build an observation
A useful observation says what was measured, where and when it happened, what source produced it, the value and units, and enough provenance to understand where the record came from.
{
"where": "47.6062,-122.3321",
"when": "",
"what": "",
"source": "",
"value": "",
"units": "",
"provenance": ""
}What to watch for
Do not overwrite source values during normalization without preserving the original meaning or recording the transformation.
Where this takes you
Provenance matters in research, standards mapping, field data, AI-assisted workflows, migrations, current-state replication, and audit.