Vendor Sales Changed. First Decide Which Clock You Mean

Two correct numbers can point in opposite directions

Take synthetic completed weeks: ordered value is $10,000 and then $9,000; shipped value is $8,000 and then $9,500. Ordered value fell $1,000. Shipped value rose $1,500. They measure different events, so neither number cancels the other.

For a vendor account in 1stPage, have Claude request vendor_analytics ordered_revenue and shipped_revenue for the same marketplace, dates and distributor view. Those are separate measures even though they arrive through one MCP connection. If the account does not expose a vendor dataset, the right next step is checking authorization and the source view, not switching to a seller sessions chart.

This illustration does not prove that a backlog cleared, a promotion ended or Amazon changed its purchase orders. Reconcile each measure against its own source total before assigning an explanation.

Amazon's Vendor Analytics manufacturing view documents orderedUnitsWithRevenue and netOrderedGMS, while shippedOrders includes shippedUnitsWithRevenue. The sourcing view documents shipped units and revenue but not the manufacturing-only ordered fields in that dataset guide. Ask whether the business question concerns retail demand, customer shipments or what the vendor receives from Amazon. Do not merge unlike measures into one 'sales' total. A purchase order from Amazon is not a customer order, either.

A null Tuesday changes the whole week

The vendor Data Kiosk guide says weekly ranges adjust to Sunday starts and completed period ends; recent requests can return null for days that are not yet available. A null is not a measured zero. Dropping Tuesday from the new week while retaining all seven days in the old one also changes the denominator. Lay out daily values, exact source boundaries and report readiness before comparing totals. Report-specific availability windows vary, so use the actual dataset and completion state rather than a blanket promise that yesterday is ready.

Inventory is another tempting bad sum: seven daily sellable-on-hand values are seven snapshots, often of the same units, not seven days of sales. Catalog attributes may reflect the current definition rather than a title as it stood in the historical sales week. Keep date basis, time zone and grouping keys beside a chart when assortment or catalog mapping changed. Otherwise a clean visual can hide a bad join.

Sourcing and manufacturing are different permissions

Amazon's Vendor Analytics Data Kiosk guide separates sourcingView, covering ASINs sourced directly from the vendor to Amazon, from manufacturingView, covering ASINs manufactured by the vendor regardless of sourcing path. The account's role determines which fields it can request. Some manufacturing measures, including glanceViews, are absent from sourcingView. A manufacturing authorization error is a missing right, not a zero-traffic observation. Write the vendor account or group, marketplace, currency, view, ASIN or brand scope, metric, grain and retrieval time above the comparison. Keep both weeks in the same view. Mixing a distributor's sourced set with a manufacturer's broader set produces a movement that might only be a coverage change. The Reports API serves both seller and vendor users, but that shared transport does not turn seller Sales and Traffic fields or seller sessions into vendor measures.

A low on-hand figure is a question, not a history

For a shipped-value decline, find the ASINs contributing the dollars first. SourcingView documents sellableOnHandInventory, sellableInTransitInventory, openPurchaseOrderQuantity and receivedFillRate. One is a stock snapshot, one concerns units in transit, one is an open purchasing commitment and one reflects receipt performance. None alone tells you how many hours an ASIN was unavailable last week. Look for dated PO and availability evidence at a compatible ASIN and period grain; mark historical availability unknown when it is not supplied.

With manufacturing rights, glanceViews can illuminate traffic, but they are not seller sessions. With sourcingView alone, do not manufacture a conversion rate from missing traffic. Confirmation rate, lead time and unfilled ordered units can inform a replenishment inquiry when the account has those measures at compatible grain. A current low inventory value plus falling shipments merits investigation, not an assertion that Amazon ordered less.

The finding should survive a change of analyst

Keep a compact evidence register with the view and permission, named metric, two completed period totals, ASIN deltas, dated availability observations and outstanding source. A finding such as 'shipped value fell for these ASINs in sourcingView, and current sellable stock is low' can be checked. 'Amazon cut purchase orders' requires dated PO records. Check an absent ASIN against assortment coverage before turning its missing row into zero revenue.

Have Claude compute deltas and propose the next source check, then verify a subtotal in the originating vendor report. In 1stPage, compare the vendor_analytics datasets this account can access and verify a subtotal in each returned source. If one view is absent, leave that branch unanswered; a seller metric cannot replace it.