Industry field guide
Media buying deal and delivery reconciliation data for AI
A media buyer may have a booking, a delivery report and a reconciliation based on different reporting windows.
- 2022 NAICS
- 541830
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
What full-period delivery is supported, and should the subset report be added or substituted to close the booking difference?
See the specimen and fields · Work through the fictional records
Task and source context
A media buyer may have a booking, a delivery report and a reconciliation based on different reporting windows. Google’s Authorized Buyers documentation describes programmatic guaranteed deals. A buyer-side exercise can test how software connects records and handles an unresolved delivery discrepancy. Google Authorized Buyers programmatic guaranteed deals
Write a fictional placement with a committed quantity and a dated delivery log. Include a later adjustment and a second report covering only part of the period. Ask the model to identify the verified period, reconcile matching records and flag the remaining discrepancy. Score contract quantity and reported delivery separately.
Use original numbers and schemas. Keep actual advertiser budgets, negotiated rates, audience identifiers, account IDs and platform exports out.
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.
541830 · Media Buying Agencies
This industry comprises establishments primarily engaged in purchasing advertising time or space from media outlets and reselling it to advertising agencies or individual companies directly.
Build a pilot sample
Set up this pilot
Use synthetic full-period and partial-period placement reports. Correct one subset with a targeted adjustment, then reconcile delivery without adding the same impressions twice.
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
- Placement and reporting window
- Defines comparable delivery scope.
- Commitment versus observed delivery
- Separates promised and reported quantities.
- Adjustment target and subset relation
- Makes the current total and overlap explicit.
Use Build an AI evaluation from professional work to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A partial-window delivery report cannot replace the full booking-period record
Original fictional placement records use abstract delivery units. They include no advertiser, negotiated rate, audience, account ID or live platform export.
| Record ID | Source | What it records |
|---|---|---|
| MB1 | Practice booking | Placement P commits 100 units over days 1–10. |
| MB2 | Reviewed full-period delivery | P days 1–10: 90 units. Authorized adjustment adds 5 to that same report, current total 95. |
| MB3 | Other delivery report | P days 6–10: 60 units, explicitly a subset of MB2's period. No evidence for the remaining five-unit booking difference. |
Question for the model. What full-period delivery is supported, and should the subset report be added or substituted to close the booking difference?
Show the reference answer
MB2 supports 95 adjusted delivery units for P over days 1–10. Against MB1's committed 100, a five-unit difference remains unevidenced in this packet. MB3's 60 covers days 6–10 already inside the full-period scope, so adding it would double-count and substituting it would shorten coverage. Keep commitment, delivery, adjustment and reporting window separate without inventing reconciliation or a payment outcome.
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
- Adding overlapping partial-period delivery to a full-period total.
- Treating a commitment as evidence that the missing delivered quantity occurred.
Change the inputs and run it again
Provide a nonoverlapping day-11–12 report outside the booking window. Require it to remain separate rather than using it automatically to fill the original five-unit gap.
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 quantities and schemas. Exclude budgets, rates, audiences, account IDs and actual platform records.
Use original or licensed practice material. Leave client work product, confidentiality obligations and regulated outcomes out of the sample.
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
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
Invisible Technologies review for expert-led AI training
Compare expert calibration and reviewer agreement before commissioning a professional-work evaluation.
Compare the approach
Mercor review for expert AI training work
Read how an individual expert role differs from licensing an employer or client’s work product.
Related industry examples
Industry guide · NAICS 541860
Direct mail creative and bundle preparation data for AI
A mailing order can specify an updated mailpiece while the bundle worksheet still references an older revision.
Industry guide · NAICS 541850
Display advertising face and installation evidence for AI
A display structure can have two faces with different creative and installation dates.
Industry guide · NAICS 541870
Flyer and sample distribution handoff data for AI
A distributor can test whether software keeps packed quantities separate from assigned packets and returns.