Skip to content

Industry field guide

Dairy milking sensor records for AI event review

A blank reading can mean the sensor stopped reporting.

2022 NAICS
112120
Updated

What this guide tests

How should minute 2 be represented in each session, and what may the assistant conclude about animal health?

See the specimen and fields · Work through the fictional records

Task and source context

A blank reading can mean the sensor stopped reporting. It should never silently become a zero. USDA’s Economic Research Service has studied precision dairy and robotic milking in the United States. A farm can demonstrate an event-review dataset with mock time series and a verified record of what happened during the gap. USDA ERS precision dairy and robotic milking research

Invent two sessions with similar curves. Make one show a confirmed equipment dropout alongside a maintenance note. Give the other a low but confirmed sensor reading and a record that sends the case to staff. Include sample intervals, units, and anonymized session keys. Test whether a model distinguishes missing measurement from a measured value and correctly routes the unresolved case. Do not make animal health findings the model’s output.

Milk processor, equipment-platform, or third-party testing contracts may govern data access. Keep actual animal records, herd productivity, staff details, and commercial terms out of the first packet.

Describe a milking data sample

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.

112120 · Dairy Cattle and Milk Production

This industry comprises establishments primarily engaged in milking dairy cattle.

U.S. Census Bureau, 2022 NAICS Manual, p. 89 ↗

Build a pilot sample

Set up this pilot

Generate two fictional sensor traces with a blank, a zero and a maintenance marker. Label the missingness and affected session without assigning an animal-health outcome.

Prepare the test packets

  1. 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.
  2. 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.
  3. 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

Session, sample interval and unit
Defines the time series.
Value and missingness flag
Preserves the measurement state.
Maintenance or validation source and referral
Supports distinct review paths.

Use Plan a farm, forest or fishery AI data sample to turn the specimen into a brief with a defined task, source and review owner.

Fictional exercise

A sensor dropout and a confirmed low measurement need different states

Two invented session traces use arbitrary measurement units and no animal identity. They illustrate record interpretation without animal-health assessment or live equipment access.

Records supplied to the evaluator
Record IDSourceWhat it records
DS1Session AAt minutes 1, 2, 3: values 4, missing, 4 U. Maintenance note links the minute-2 gap to a confirmed reporting dropout.
DS2Session BAt minutes 1, 2, 3: values 4, 0.2, 4 U. Separate validation card confirms the 0.2-U reading; no dropout flag.
DS3Review ruleMissing values remain missing. Valid low measurements go to staff review with their source; the exercise supplies no health rule or outcome.

Question for the model. How should minute 2 be represented in each session, and what may the assistant conclude about animal health?

Show the reference answer

DS1's minute-2 value is missing with a documented sensor-reporting dropout, not a zero measurement. DS2's 0.2-U value is a confirmed low reading and should retain that number and its validation source. DS3 calls for staff review of the latter, but provides no basis for a health conclusion in either case. Similar curve shapes do not make the missing and measured states interchangeable.

Download the fictional records and answer key (JSON)

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

  • Filling a missing sensor interval with zero.
  • Discarding a validated low measurement as a dropout or turning it into a health diagnosis.

Change the inputs and run it again

Remove A's maintenance note. Require the value to stay missing while its cause becomes unknown, rather than retaining the prior dropout explanation.

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 invented traces. Exclude actual animal records, productivity, staff and commercial terms; processor and equipment-platform contracts may control real exports.

Use owner-approved administrative records or staged material. Exclude unnecessary animal, farm, staff and customer identifiers.

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

How the examples work

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.

All company reviews

Related industry examples

← Agriculture, forestry and fishing