Amazon settlement reports are where the finance side of your Amazon business gets real. Every charge, every refund, every fee adjustment, every disbursement — it all flows through them. They are also where most accounting teams give up because the data is dense, marketplace-specific and not built for human reading.

This guide explains how to actually use them.

TL;DR: Amazon settlement reports are the source of truth for what Amazon paid you and what it took from you. They list every transaction at the line level — sales, refunds, fees, adjustments, reserves — grouped into settlement periods (typically two-week cycles). Reconciling them against your finance system requires joining settlement lines to orders by transaction reference, applying time zones consistently, and handling reserves correctly. Done well, you have a daily-fresh net revenue number with no spreadsheet work.

What a settlement report actually contains

Each settlement covers a defined period and lists, line by line:

• Order revenue — gross sale, by order ID and SKU.
• Promotional discounts — amounts deducted from gross.
• Refunds — returns processed during the period.
• Amazon fees — referral, FBA fulfillment, storage, removal, advertising.
• Reimbursements — lost, damaged or warehouse-error reimbursements.
• Adjustments — chargebacks, FBA fee corrections, Amazon-issued credits.
• Reserves — amounts held by Amazon (account reserve).
• Disbursement — the net amount actually paid out.

Each line has a timestamp, an order reference (where applicable), a marketplace and a currency.

The reconciliation problem

The thing that makes settlement reports tricky is that the same order can appear across multiple settlement periods. A sale in Period A, a refund in Period B, a fee adjustment in Period C. To reconstruct order-level economics, you have to join across periods.

Add multiple marketplaces, multiple currencies and the standard 14-day settlement cycle, and a single seller with five active marketplaces is dealing with a lot of moving pieces every two weeks.

How to reconcile properly

Step 1: Ingest every settlement

Pull settlements from every active marketplace as they post. Store at line-level granularity, not aggregated.

Step 2: Join lines to orders

Use Amazon order ID as the foreign key. Each order will likely have multiple settlement lines (sale, fee, refund) sometimes spread across periods.

Step 3: Apply currency and time zones

Convert every line to your reporting currency using period-appropriate FX rates. Apply marketplace-local time zones for daily metrics, your global time zone for executive reports.

Step 4: Match against finance system

Compare aggregated disbursements against bank deposits and your accounting system. Variances over a defined threshold (often $500 or 1%) get flagged for review.

Step 5: Track reserves separately

Reserves are revenue you have earned but not received. They affect cash flow but not P&L recognition. Most teams ignore them and end up confused when disbursement timing shifts.

Common reconciliation issues

• Refund timing. Refunds can lag the original sale by 30+ days, crossing settlement periods.
• Fee adjustments. Amazon corrects FBA fees retroactively. The adjustment shows up later than the original fee.
• Reimbursements. Lost-inventory reimbursements may not be tied to a specific order — they need their own bucket.
• Multi-currency rounding. Currency conversion at the line level introduces tiny rounding errors that compound across thousands of lines.

How AI builders make this easier

The reconciliation logic is rule-based, not creative. Once your data layer has clean settlement lines, an AI builder can ship the full reconciliation pipeline as a one-prompt project:

“Build me a daily reconciliation job that compares Amazon disbursements against QuickBooks deposits per marketplace per currency. Flag any variance over $500 or 1% and post to #finance.”

The AI handles the joins, the FX conversion, the threshold logic. Your finance team gets a clean dashboard.

The bottom line

Settlement reports are dense by design. The reconciliation work is rule-based and repetitive — exactly the kind of work that should run on a data layer plus an AI builder, not in a spreadsheet on someone’s laptop.

DataDoe’s Amazon data layer ingests settlements at the line level across every connected marketplace and exposes them through SQL, API and MCP for AI builders.