Skip to main content
All case studies

Reconciling missing inventory with reimbursements

For an unnamed kitchen-tools brand, the question was which inbound inventory shortages had already been accounted for and which still needed investigation. A short shipment alone could not answer it: a payment might exist, or extra receipts elsewhere might change the reconciliation.

The DataBrill team’s historical analysis compared Fulfillment by Amazon (FBA) shipment records with receipts and reimbursements. It matched records by stock keeping unit (SKU), the product identifier used in the reconciliation. That made the report’s limits part of the answer from the start.

Start with the shipment cohort

The amended analysis covered 281 closed shipments with item records, destined for the US and created from April 1 through July 31, 2026. Of those, 92 shipments arrived short, with a total shortage of 1,000 units counted across their short lines.

Account for the 1,000 short units
  1. 440 paidReimbursements matched by SKU
  2. 294 offsetOver-receipts of the same SKU
  3. 266 unresolvedStill needs investigation
Left to right: 440 paid + 294 offset + 266 unresolved = 1,000 units. US closed shipments created April 1–July 31, 2026: 281 shipments, 92 short. Matching is by SKU, not shipment; offsets do not trace physical units. Unresolved does not mean proven owed. Payment approvals from April 20 onward were reviewed in October 2026 and capped at each SKU’s shortage.

This is a shipment-creation cohort, not a group selected by the date Amazon received the goods or approved a reimbursement. Preserving that definition makes it possible to revisit the result without silently changing which shipments count.

Two inbound plans with no receipts were treated as not shipped and excluded. That was an assumption used to set the analysis boundary; counting them as lost inventory would have added a claim the records did not establish.

Reconcile the short units

Outcome Units Meaning
Paid 440 Reimbursements matched by SKU
Offset by over-receipts 294 Extra units of the same SKU received in other shipments in the window
Unresolved 266 Neither matched payment nor an offset in this analysis
Total short 1,000 Starting shortage across short shipment lines

US closed shipments created April 1-July 31, 2026. The analysis matched Lost_Inbound reimbursement approvals from April 20 onward in the records reviewed in October 2026. Payments counted against each SKU were capped at its shortage.

The reimbursement report contains no shipment ID, so a SKU match does not prove that a particular shipment was reimbursed. The table is a reconciliation across records with different levels of detail. Its payment total should not be read as a shipment-by-shipment audit trail.

The over-receipt category needs similar care. If another shipment of the same SKU received more than was recorded as sent, the analysis used that excess to offset shortages. This is an accounting treatment for the comparison, not proof that the original physical units were traced into another shipment.

After those treatments, 266 units remained unresolved. That means they needed investigation, not that Amazon had been proved to owe reimbursement for all of them.

Keep later reversals in view

A payment can also change after the first reconciliation. In a separate US inbound-payment cohort, $1,763 of $9,514 paid during January-September 2026 was later reversed: 18.5% of those dollars.

That denominator belongs beside the percentage. This is a dollar-payment cohort from a different period, not a finding that 18.5% of the 1,000 short units were reversed. Combining the two would hide both the time window and the kind of quantity being measured.

For a repeatable review, retain the original payment and its later adjustments together. A list of approvals alone can give a different picture from the net position after reversals.

Turn the remainder into an investigation list

The next task is to inspect the unresolved items against shipment and receipt history, same-SKU payments, later adjustments, and supporting case records. A mismatch may require information outside the reports already collected. The database cannot supply a missing shipment reference that Amazon’s reimbursement report never contained.

Keep the matching rules and exclusions with the result. Another reviewer should be able to tell why a unit is in the unresolved category and which additional record could change that classification. This produces a narrower investigation list than treating every shortage as a fresh claim.

What this analysis establishes

Claims were already being filed before the database, using downloaded reports. This comparison does not establish incremental recovery from Core or a verified improvement in profit. Its practical outcome is clearer reconciliation and a more focused set of records to investigate. It also makes no claim about today’s reimbursement policy or a particular item’s current eligibility.

From Amazon reports to a database your team can use follows the workflow from collection to a reviewable answer. Why did your Amazon sales change? shows why a different business question needs different records. For the pipeline and owned-data layer, see Core’s platform.