Skip to content

INDUSTRY FIELD GUIDE

Farmed shellfish tag and harvest-lot data for AI

Two harvest tags can look alike and still describe different lots.

By Fiund · Updated 2026-11-05

NAICS 112512 · Shellfish Farming

Two harvest tags can look alike and still describe different lots. FDA’s National Shellfish Sanitation Program materials provide the context for molluscan shellfish records, and NOAA describes shellfish aquaculture operations. A farm can use mock tags to demonstrate how it keeps lot relationships and verifiable evidence separate. FDA NSSP guide · NOAA shellfish aquaculture overview

Use three generic tags with clearly fictional areas and dates. Give one a verified association to an invented harvest lot and give a second a similar area name from a different day. Record a documented split with two child IDs. The evaluator can ask the model which tags it can trace to the original lot and where a mismatched date prevents a match. Keep any public health or release decision outside the task.

Do not share lease boundaries, farm security details, vendor routes, or current harvest permissions. Use dummy tags and ask all organizations controlling real records before using them.

Describe a shellfish lot sample

Original Fiund exercise

A documented lot split preserves ancestry while a similar tag date does not

Original dummy tags use fictional area labels, dates and lots. They are not official tags, harvest permissions or a food-safety release record.

Fictional records supplied to the evaluator
RecordSourceWhat it records
SHF1Origin tagTag T1 links lot L1 to practice area A on day 2.
SHF2Split recordAuthorized fictional split links L1 to child lots L1-A and L1-B on day 3. Child tags explicitly cite T1 and the split reference.
SHF3Similar tagTag T2 says area A, day 4, lot L2. No parent or split relationship to L1 supplied.

Question for the model

Which child tags can be traced to the day-2 origin lot, and does T2's matching area establish the same ancestry?

Show the reference answer and checks

SHF1 and SHF2 explicitly connect L1-A and L1-B through the split record to T1's day-2 L1 origin. SHF3 describes another lot and day with no parent link, so its shared area label does not establish L1 ancestry. Preserve date, lot and split relationships together. The trace exercise establishes only the supplied documentary links, not current harvest authorization, product safety or permission to release a lot.

  • Joining lots by area name while ignoring harvest day and lot identity.
  • Treating a reconstructed lineage as a public-health or release decision.

Fields to preserve

Tag, area alias and harvest date
Defines the origin reference.
Parent and child lot IDs
Preserves the split.
Verified link and missing relationship
Separates supported ancestry from resemblance.

Test another case

Remove L1-B's parent reference while retaining its similar label. Require that child's trace to become unresolved rather than inferred from naming.

Use original dummy tags. Exclude leases, routes, security details and current permissions; real records require authorization from every controlling organization.

Download this fictional example and answer key (JSON)

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.

Primary sources and further reading