Tracking that stops moving: what to check, in order

Seven checks, scored before any of them are ranked

Ranking checks is only honest if the scoring happens before the order is chosen, because a list assembled from anecdotes will always put the writer favourite first. The desk scores each check on two things: how often it has ended a stall in the case files reviewed, and how much it costs to run. The second part matters because a check that takes four days to answer should not sit above one that takes ten minutes.

Cost is measured in days rather than in money, since almost every check on this list is free to run. A check that only requires reading a page costs nothing and completes immediately. A check that requires a reply from another party costs at least one working day, and usually two, which is why the cheap checks have to be exhausted first even when the expensive one feels more decisive.

Two checks were left off the list on purpose. Refreshing the tracking page more often is not a check, and contacting the seller about a parcel that has already been handed over is not one either, because the seller stopped being able to affect the parcel the moment the warehouse took it. Both feel like action and neither changes any input.

The ranking, and the weights behind it

The weights below add to one hundred and express how much of the decision each check resolves, not how interesting it is. A check weighted twenty closes a fifth of the question for the reader on its own. A check weighted ten narrows the possibilities without ending them, and the two at the bottom of the list are the ones that generate paperwork rather than answers.

The order is also the order of cost. The top three are all readable from records the reader already holds: the scan history, the lane window, and the receipt for what was handed to the line. The bottom four each require another party to answer, which is why they appear last even where they eventually matter most.

Checks one to three: the ones that close a case

The scan event that appears to be the last one is frequently not the last one in substance. A handover scan, a build scan and a departure scan can all read as a place name, and only one of them means the parcel is physically moving. Reading the event type rather than the location is what turns a nine-day silence into either a normal build wait or a genuine gap, and the two lead to completely different next moves.

Matching elapsed days against the lane window is the second check, and it requires the window to be a range rather than a number. On the desk lanes the United States air-economy window runs from 8 to 16 days on 214 reported arrivals, while the United States sea window runs from 25 to 42 days on only 38 reports. A parcel on day 12 is inside the first window and nowhere near the second, and treating those two as the same kind of late is how readers end up chasing a parcel that is simply on a slow mode.

The third check is the one most readers have the evidence for and never look at: what the warehouse recorded as handed to the line, on what date, under which tracking number. A discrepancy there explains a surprising number of stalls, because a parcel can be scanned into a build under a number the buyer never received, and a tracking page showing nothing is then perfectly accurate rather than broken.

Checks four to seven: the ones that buy information

Splitting and merging is the fourth check and the most underrated one. A parcel that was consolidated twice can keep its original tracking number while changing shape, and a build that was re-cut to catch a weekly cut-off can leave a gap of a week with no scan at all. The question to ask is not where the parcel is but what happened to the build it was in.

The fifth and sixth checks are about position and confidence. Knowing which side of the border the parcel is sitting on tells you which queue it is in, and the queues behave differently: a customs clearance queue and a biosecurity queue in the same destination can add different amounts of time, and a member-state clearance inside the European Union can take two days for one parcel and ten for an identical one. Knowing that a lane has a thin sample tells you to widen the expected window rather than to escalate, and the desk labels any lane under 45 reported arrivals as thin for exactly that reason.

The seventh check is the formal trace, and it sits last because it is the only one that spends something you cannot get back. Opening a trace starts a clock on the claim side, and a claim opened while the parcel is still inside its window will usually be answered with the window itself. The trace is the right move after the free checks have been exhausted and the window has been passed, and the wrong move before.

Where this ranking is open to argument

The obvious objection is that a reader whose parcel has already passed its window should start at check seven and ignore the rest. That is fair, and the ranking is written for the reader who does not yet know which situation they are in. Once the window has clearly been passed, the ranking collapses and the claim path takes over, because the checks above it can no longer change the outcome.

A second objection is that sea and rail should be ranked differently rather than sharing a list with air. There is something in this: on both of those modes tracking updates in blocks rather than step by step, so a long silence carries much less information, and the useful question shifts from where is it to which block is it booked into. A reader on a sea build who runs the first three checks is mostly confirming that nothing is wrong.

A third objection concerns the weights themselves. They are a desk convention, not a measurement, and a reader with a different order of priorities will weigh them differently. What the desk will defend is the cost ordering rather than the exact figures: free checks before paid ones, records before replies, and replies before anything that starts a clock.

Running the list against a live tracking number

Run against a real number, the list takes about ten minutes for the first three checks and produces one of four readings: the parcel is inside its window and normal, the parcel is outside its window on a well-reported lane, the parcel is outside its window on a thin lane, or the record itself does not agree with what was handed over. The last reading is the only one that justifies contacting two parties at once.

Whichever reading comes out, the same three things should be written down before any message is sent: the last scan event and its date, the lane window with the sample count beside it, and the date the warehouse recorded the handover. Those three lines are what makes a question answerable in one pass, and they are also what a claim will be assessed against later if the parcel does not turn up.

What the list deliberately does not do is predict an arrival date. The desk records windows as the outer bounds of reported arrivals, so the honest output of this exercise is a range and a next action, not a day. A reader who wants a day will get one from somebody, and it will be a guess with a confident tone.

The data point behind this note

All thirteen lanes were re-cut during batch KB-2026-06 with an explicit sample count per lane, ranging from 214 reported arrivals on the United States air-economy lane down to 29 on New Zealand, and lanes under 45 reports are labelled thin on the gauge rather than presented as settled.

◇KB-2026-06· 2026-09-30 Batch ledger · Challenge a figure

Notes in the same cluster

Related entries and gauges

• How do you trace a parcel whose tracking stopped moving?
• How do you file a missing-item claim that gets answered?
• Transit window bench

  Reported arrival windows with the sample count attached.