Industry field guide
Commodity broker transaction corrections for AI evaluation
A corrected transaction record may replace a field without creating another trade.
- Sector
- Finance and insurance
- 2022 NAICS
- 523160
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
How many distinct identified transactions are present, which field changed for T-A, and can BC3 be applied to either transaction?
See the specimen and fields · Work through the fictional records
Task and source context
A corrected transaction record may replace a field without creating another trade. CFTC materials describe how swap data reporting errors and corrections are handled in one part of the derivatives market. A broker can build a mock sequence to evaluate whether a model preserves the original transaction identity and corrected version. CFTC swap data correction advisory
Invent two order and transaction aliases, an original confirmation, and a later replacement with a clear cross-reference. Give an unrelated transaction the same quantity. Ask which recorded field changed and which transactions remain distinct. Add a missing link so the model must flag uncertainty. A trained reviewer should decide the expected answer and exclude any trading recommendation from the rubric.
Keep positions, clients, prices and strategies, trade venue IDs, execution logs, and access credentials private. Check exchange or platform rights before reusing actual formats or files.
Industry research
The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.
523160 · Commodity Contracts Intermediation
This industry comprises establishments primarily engaged in acting as principals (i.e., investors who buy or sell for their own account) in buying or selling spot or futures commodity contracts or options, such as precious metals, foreign currency, oil, or agricultural products, generally on a spread basis, or acting as agents (i.e., brokers) in buying or selling spot or futures commodity contracts or options on a commission or transaction fee basis.
Build a pilot sample
Set up this pilot
Build two mock trade confirmations and a correction targeting only one transaction field. Add an unmatched note with a similar order number and grade both cardinality and version lineage.
Prepare the test packets
- Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
- Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
- Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.
Fields the reviewer needs
- Transaction and order aliases
- Disambiguates equal-quantity records.
- Version and supersedes reference
- Defines the replacement chain.
- Changed field and unmatched-note status
- Separates an evidenced correction from an assumption.
Use How to prepare a financial workflow data sample for AI to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A replacement confirmation changes a field without adding a trade
Imaginary transaction aliases and arbitrary quantity units create an offline correction fixture. No security, contract price, position or strategy is represented.
| Record ID | Source | What it records |
|---|---|---|
| BC1 | Confirmation V1 | Transaction T-A, order O-A, quantity 12 units, category X. |
| BC2 | Replacement V2 | Explicitly supersedes BC1 for T-A/O-A; category corrected to Y. Quantity remains 12. Separate T-B/O-B also has quantity 12. |
| BC3 | Unlinked correction note | Says category Z and quantity 12 but omits transaction, order and supersession references. |
Question for the model. How many distinct identified transactions are present, which field changed for T-A, and can BC3 be applied to either transaction?
Show the reference answer
BC1 and BC2 describe two identified transactions, T-A and T-B; the replacement is a new version of T-A, not a third transaction. T-A's category changes from X to Y while its quantity stays 12. BC3 lacks the identifiers needed to associate its Z correction with either transaction. Equal quantity cannot resolve that gap, so retain the note as an unmatched exception for review.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Counting a replacement confirmation as another trade in totals.
- Applying an unlinked correction to whichever transaction shares its quantity.
Change the inputs and run it again
Give BC3 an explicit T-B reference but a conflicting order alias. Require a flagged identity conflict rather than a silent update.
Score the decision and the evidence separately
Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.
Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.
Check permission and scope
Use original dummy records only. Exclude execution logs, clients, venue credentials and licensed platform formats; no trading action or recommendation belongs in the task.
Use public or fabricated examples only. Omit customer holdings, access credentials, actual trading instructions and advice.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Includes worksheet
How to prepare a financial workflow data sample for AI
Build a fictional or authorized finance example with the reporting unit and time cutoff intact.
Includes worksheet
Plan a sample of business data for AI
Describe the evidence without transmitting private account or customer information.
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Compare the approach
Scale AI review: data partnerships, Outlier and buyer services
Read the public financial-services partnership criteria and scope controls.
Compare the approach
Invisible Technologies review for expert-led AI training
Compare expert calibration and reviewer agreement for narrow financial document exercises.
Related industry examples
Industry guide · NAICS 523150
Investment banking filing amendments for AI review
An amended filing may change one section while leaving the rest of an earlier filing relevant.
Industry guide · NAICS 523940
Investment advisory fee schedule data for AI evaluation
A public brochure may describe fee categories while a particular account agreement documents the arrangement used for a statement.
Industry guide · NAICS 523999
Securities clearing and settlement status data for AI
A transaction may be matched or accepted for clearing without a settlement confirmation.