Skip to content

INDUSTRY FIELD GUIDE

Human services program caseload definitions and reporting data for AI

A benefits dashboard can count participating families while another report counts individuals.

By Fiund · Updated 2026-11-05

NAICS 923130 · Administration of Human Resource Programs (except Education, Public Health, and Veterans' Affairs Programs)

A benefits dashboard can count participating families while another report counts individuals. HHS’s TANF data resources provide public reporting context. A program office can create an original exercise that tests whether a model compares the correct measure without handling anyone’s benefit file. HHS TANF data portal

Use two mock monthly summaries with different counting units. Supply an explanatory note and one total revised after publication. Ask for the supported comparison and require the answer to cite the unit, month and version. Keep the family and person figures separate.

Exclude applicants, benefit histories, incomes, household details and eligibility decisions. Use fictional numbers or explicitly released aggregates with permission and applicable disclosure checks.

Describe a fictional caseload reporting sample

Original Fiund exercise

Families and individuals are different caseload counting units

A dummy benefits dashboard contains invented monthly aggregates. The task makes no eligibility, payment, income or household-level decision.

Fictional records supplied to the evaluator
RecordSourceWhat it records
BG1Month M family report revision 140 participating families; unit explicitly families.
BG2Authorized family revision 2Corrects M to 42 families. Individual report for M lists 100 participating people.
BG3Month N family report45 families, same stated family definition as M. No N individual count supplied.

Question for the model

Which month-to-month family comparison is supported, and can 100 individuals replace M's family count or establish N's person count?

Show the reference answer and checks

BG2 makes 42 the current Month M family figure, while BG3 reports 45 families under the same stated definition for N, a difference of three families. M's 100 participating individuals is a separate counting unit and cannot replace 42 families. No N person count is supplied, so a person-level trend is unresolved. Preserve metric unit, month and revision in both the reported values and comparison.

  • Subtracting a person count from a family count as a caseload trend.
  • Using the superseded M value of 40 or assuming a fixed people-per-family conversion.

Fields to preserve

Counting unit and definition
Separates families from individuals.
Reporting month
Defines compatible comparison periods.
Revision authority
Identifies the current published aggregate.

Test another case

Change N's family definition while leaving its value 45. Require a comparability warning rather than repeating the earlier three-family interpretation without checking metadata.

Use synthetic aggregates. Exclude applicants, benefit histories, incomes, actual household details and eligibility records.

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