Research question and scope
This protocol measures how often a fulfillment desk ships a duplicate parcel after an address edit. The unit is one order with at least one address edit and at least one label event. A duplicate means a second parcel left the building for the same order without a recorded reship reason. The measure describes the rate of accidental repeats after a change, not the total parcel volume.
Set the scope to one warehouse, one client, and a fixed period. Record the order system, the label system, and the carrier feed used to detect parcel events. If the warehouse changed its process during the period, split the sample at the effective date. A different process creates a different denominator.
Keep the measure separate from the customer's intent. A customer may ask for a replacement after a genuine loss, and that second parcel is not an error. The protocol counts only parcels without a recorded legitimate reason. Report legitimate reships separately so the reader can see both.
State the decision the result will support before collection begins. The client may use it to add a label hold, change the address edit workflow, or require approval for a reship. Write the intended action beside the scope. A measure with no decision becomes a dashboard number.
Population and unit of analysis
Define the population as every order that had an address edit during the period. Include edits made by the customer, the support desk, and the warehouse. The unit is the order, not the edit or the parcel. An order edited twice is still one order unless it produced two separate shipments.
Exclude orders that were cancelled before any label was purchased. Record the excluded count and the reason. A cancelled order cannot produce a duplicate parcel. Report the exclusion so the reader can judge the active population.
Include orders handled through any channel, including manual warehouse corrections and bulk uploads. Record the channel and the role that made the edit. Leaving out a bulk path hides a common source of repeats. State the inclusion rule before drawing the sample.
Decide how to treat a parcel that was shipped, returned, and then shipped again. Score the second shipment only when it was tied to the same order and no reship reason was recorded. Record the return tracking and the approval. A return can make a second shipment legitimate.
Definitions that make the measure reproducible
Define an address edit as any change to street, unit, city, postal code, or country after the order was placed. Record the old value, the new value, the time, and the source. A phone number change is not an address edit. The field set should be specific enough to reproduce the count.
Define a duplicate parcel as a second label with a pickup or carrier acceptance event for the same order, with no recorded reship reason. Define a legitimate reship as a second parcel with a reason, an approver, and a date. Record both definitions in the scoring guide before the sample is drawn. Give one example of each before sampling.
Define the observation window for the second parcel. A repeat may ship on the same day or several days later. Record the window, such as fourteen days after the edit. The window should be long enough to catch slow repeats and short enough to keep the sample current.
Define an address edit that was never saved. A specialist may correct a field in a carrier portal without updating the order. Treat that as an edit only when the carrier record shows it. Record the source used. A silent portal edit is still part of the customer experience.
Sampling and data collection
Draw the sample from the order system's edit log for the full period. Join the edit records to label and parcel events using the order identifier. Record the export date, the join rules, and the row counts at each step. A join that drops rows will understate the rate.
Capture the order identifier, the edit time, the old and new address, the label time, the parcel scan, the channel, and the role. Add the reship reason and approver when one exists. Store the joined extract with the report. The join is the core evidence.
Check a small overlap sample by reading the order history by hand. If the join misses a manual correction or a portal edit, revise the join before the full sample. Record the discrepancy. A clean join is more important than a large sample.
Record the time spent joining and reading each case. A measure that requires three exports and a manual review will not be repeated. Note whether the order system exposes a stable edit log. Collection cost belongs beside the result.
Measuring duplicate shipments
Report the share of edited orders that produced a duplicate parcel. Show the numerator and denominator beside the share. A percentage without the order count is hard to interpret. The share should describe edited orders, not all orders.
Break the result down by edit source, edit timing, and warehouse shift. An edit made after label purchase may be riskier than an edit made before picking. Compare only groups with the same process version. A rate without the group size is a story.
Report the most common cause and the most common missing record. A missing void confirmation may point to the carrier workflow, while a missing reship reason may point to approval. Record one example of each. The pattern should lead to one control.
Report the cost of the duplicate parcels separately from the count. A small number of high-value repeats may matter more than a larger number of low-value samples. Record the shipping and replacement cost. Cost helps the client prioritize the fix.
Limitations and next action
The measure cannot show whether a customer received two parcels or whether one was intercepted. It also cannot show a duplicate that shipped under a different order identifier. Treat the result as a measure of recorded duplicates. Note the systems and channels that were not included.
Give the result to the fulfillment lead and the client contact with one change to test. The change might be a label hold after an address edit, a required void confirmation, or a reship approval step. Pick one change and name the owner who will apply it. Record the effective date and repeat the measure with the same definitions.
Review the boundary cases before tightening the rule. A customer who moves during a long order cycle may need a new parcel even when the old one is still active. Fix the exception list first if that is the pattern. The goal is fewer accidental repeats, not fewer shipments.
Schedule the next measurement for the same week in the following month. Keep the warehouse and channel filters unchanged so the comparison is fair. Record any process change in the note. A trend built from different rules is not a trend.
