Raw data and processed results should remain distinguishable
1 min
A final result often follows filtering, correction, calibration and aggregation. Preserving only that result removes the evidence needed to assess those transformations.
Explanation
Raw observations need their acquisition context, while processed outputs need the operations and parameters used. Neither is fully interpretable without its relationship to the other.
Example
A teaching temperature average changes after a different smoothing window is selected. Keeping raw readings allows the effect to be evaluated without repeating the physical experiment.
In practice
Retain original observations and version processing steps. Avoid replacing raw files with exported summaries that cannot support independent reconstruction.
Sources
GO FAIR Foundation — FAIR Principles: Identifiers, Metadata and Provenance