INDUSTRY FIELD GUIDE
Laboratory instrument data for AI observation-version review
NIST publishes measurement-traceability guidance.
NAICS 334516 · Analytical Laboratory Instrument Manufacturing
NIST publishes measurement-traceability guidance. A laboratory-instrument manufacturer can propose an original task around the distinction between a raw observation, a processed result and the version of the method reference supplied. NIST measurement context
Create fictional instrument identifiers, observation files and a processing register. Give one result a complete link to its raw source and processing version. Leave that link absent in another case. Ask the assistant which result answers the selected request. Include a later reprocessing record that changes the output while preserving the original source identifier.
Annotate instrument, source observation, processing version, sample reference, timestamp and status. Evaluate with a new file-naming scheme and table layout. Score raw and processed values confused, a later result treated as the original observation and an unsupported source join independently. The key should retain each version's provenance.
Use original records and invented values with no laboratory procedure supplied. Customer samples, licensed analytical methods, proprietary software and outside reports need explicit authority and qualified review. Describe the observation-history task and checked examples in the licensing brief. Confirm buyer requirements before preparing a larger archive.
Discuss this original manufacturing sample with Fiund
Original Fiund exercise
Reprocessing creates a new view of the same observation
Fictional instrument records use neutral file keys. No real lab or sample data appears.
| Record | Source | What it records |
|---|---|---|
| 1 | Observation register | Raw capture O8 recorded once by instrument I4. |
| 2 | Processing log | Outputs V1 and V2 both derived from O8; V2 current. |
| 3 | Dashboard draft | Counts V1 and V2 as two observed samples. |
Question for the model
How many raw observations should the dashboard count?
Show the reference answer and checks
One: O8 is recorded once in record 1. Record 2 produces two processing versions of that observation, so use V2 for the current view and retain V1 as history. Record 3 double-counts versions as distinct observed samples.
- Counting processed files as independent observations
- Discarding processing history when choosing the current view
Fields to preserve
- raw_capture_id
- Defines the observation grain
- processing_version
- Tracks derived outputs
- current_view
- Selects the effective version
Test another case
Include a second independent capture processed by the same version; count two raw observations with one shared processing version.
No actual lab data, methods, instrument settings, patients or proprietary files.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.