A useful China fulfillment dashboard explains where orders stop, which inventory can be sold, and what an exception costs. It should let an operator reproduce a number from individual records. A screen showing twelve green percentages without their denominators cannot do that.
Start with one agreed reporting boundary. An order received from a store, an order released to the warehouse, a packed parcel and a carrier-accepted parcel are different events. A dashboard should preserve those distinctions rather than label all four “shipped.”
Define twelve metrics before choosing the charts
The following is a proposed operating dictionary, not a set of FulfillNexa performance claims or industry targets.
| Metric | Calculation or measurement boundary |
|---|---|
| Receipt discrepancy rate | Inbound lines with a quantity or identity discrepancy divided by received lines |
| Receiving-to-available time | Time from physical receipt to approved sellable inventory |
| Inventory record accuracy | Counted SKU-location records matching the approved tolerance divided by counted records |
| Order release backlog | Accepted orders awaiting warehouse release, separated by hold reason |
| Pick accuracy | Correctly picked lines divided by inspected picked lines |
| Pack accuracy | Inspected packages matching the approved order and pack specification divided by inspected packages |
| Release-to-pack time | Elapsed time from warehouse release to pack completion |
| On-time carrier handoff | Eligible orders accepted by the carrier within their agreed window divided by eligible orders |
| Missing first-scan rate | Handed-over parcels lacking the expected carrier evidence after an agreed review interval |
| Exception age | Time since an exception opened, grouped by unresolved cause |
| Return reason rate | Returned units for a defined reason divided by delivered units in the same eligible cohort |
| Fulfillment cost per order | Included warehouse costs divided by the orders that incurred those costs |
Record whether a metric counts orders, lines, units, parcels or inventory records. One order containing four products and split into two parcels contributes differently to each denominator. Combining these measures makes apparent improvements difficult to interpret.
Separate inventory availability from physical possession
The warehouse may possess goods that are already committed, damaged, awaiting inspection or reserved for another purpose. Shopify’s inventory-state documentation distinguishes these states rather than treating on-hand stock as immediately sellable.
Apply the same discipline to the dashboard. A shortage alert should identify the relevant SKU and location, the quantity physically present, the quantity held, and the remaining saleable quantity. Otherwise a store team may keep promoting a product that appears plentiful in the warehouse but cannot be released.
Receiving speed also needs a boundary. Report both physical receiving time and time to approved availability. A supplier discrepancy can make the second measure slower without indicating slow unloading.
Use an order cohort that can be audited
Consider an illustrative day with 1,000 orders received. Fifty are awaiting customer clarification and fifty require a stock investigation. Of the remaining 900 eligible orders, 873 receive carrier acceptance within their agreed handoff window.
On-time handoff for that eligible cohort is 873 ÷ 900, or 97%. Reporting 873 ÷ 1,000 answers a different question. Neither calculation justifies hiding the 100 held orders. Display the holds beside the service result, with owners and ages, so the exclusions remain visible.
Suppose 20 of the 873 accepted orders contain two parcels. An order-level result still counts 873 orders, while parcel-level scan coverage must reconcile all 893 parcels. This is why the drilldown should retain order IDs and parcel IDs separately.
Store the evidence behind each timestamp
For every operational event, retain its subject, location, event time, receiving system and evidence reference. GS1 EPCIS provides a standard for sharing visibility events; adopting that standard is a separate integration decision, not a prerequisite for defining an honest dashboard.
A carrier scan arriving late should not overwrite the time the warehouse received the message. Both timestamps help explain whether the parcel moved late or the information arrived late. Use one declared time zone for daily reporting and preserve original offsets in the underlying records.
Make the dashboard produce work
Assign an action to each exception view. An aging stock discrepancy goes to inventory control. A missing scan goes to the transport owner. A pack-error trend goes to the person responsible for the approved packing instruction.
Before setting targets, reproduce one week of results manually from sampled orders and investigate every unexplained difference. The first useful dashboard deliverable is a signed metric dictionary and a reconciled order cohort. The chart design can follow.
Sources and verification links
Operational rules can change. Check the linked primary sources and confirm shipment-specific requirements before booking.