Forty Parcels, One Spreadsheet: What Transit Records Show
2,009 words · about 10 minutes · note 24 of 40 · Step 4
Forty parcels is a comfortable number to keep and an uncomfortable number to reason from. It is large enough to make a spreadsheet look like evidence and small enough that a handful of late arrivals can move every figure on it. The register itself is simple — dispatch date, transport product, weight, arrival, and the gap between the fastest and slowest promise — but what it can and cannot answer is not simple at all. This note follows one such register through five arguments, states the counterargument honestly, and ends with the boundary conditions under which forty rows are worth citing.
Why a transit register is kept at all
A dispatch decision has two halves. The price half is visible before anything is sent: what the goods cost, what the packed parcel weighs, what the freight line charges for that weight. The time half is invisible until after the fact, and it is the half people try to remember instead of recording. A register exists to move that second half out of memory and into a column that can be re-read in six months without reconstructing anything.
The usual content is modest. One row per parcel, with the date the seller was paid, the date the warehouse released the shipment, the transport product used, the weight the bill was based on, the arrival date, and whatever range the product published when it was chosen. Two of those fields are facts about your order; one is a promise made by somebody else. Keeping them in separate columns is the difference between a record and a rumour.
What follows uses the register format this site describes in its transit comparison tool, where a line’s published range is entered as two ends and the gap between them is treated as the primary column. Where numbers appear below, they come either from this site’s own snapshot of 218 catalogue listings or from arithmetic shown in full. Measured transit days for any named transport product are not something this site holds, and they stay unverified throughout.
Part one: the register measures handovers, not parcels
A parcel does not travel continuously. It changes custody: seller to domestic courier, courier to warehouse intake, intake to a shelf, shelf to a packing bench, bench to an export handover, then through customs, a flight or a vessel, an arrival hub, a clearance agent and a final delivery agent. Each change of custody is a place where a row of data can be dated, and none of them is the parcel itself.
That distinction explains the most common misreading of a tracking page. A status that sits unchanged for four days is usually a queue — waiting for a flight, waiting for a customs queue to clear, waiting for a delivery round — and a queue is not the same event as a loss. A register that records handover dates lets the gap be attributed to a stage instead of being blamed on the whole journey.
It also changes what a row is evidence for. A single row can tell you that this parcel was released from the warehouse on one date and delivered on another. It cannot tell you that the transport product is fast or slow, because the product controlled only part of the interval, and the seller, the warehouse and the destination clearance each controlled their own. Forty rows of the same shape do not remove that ambiguity; they only make its size visible.
Part two: forty rows describe spread, not a ranking
Averages are the wrong summary for this kind of record. Two transport products can share a twelve-day average and behave nothing alike: one arrives between day nine and day fourteen, the other between day five and day twenty-four. The second is not worse on average; it is simply less predictable, and predictability is what a deadline needs. This site’s comparison tool therefore treats a gap above ten days as an uncertain window, a threshold that belongs to this site rather than to any carrier.
Small samples make the point sharper. With six observations, a median is the average of the third and fourth values once the rows are sorted. Six illustrative arrivals of 9, 11, 12, 14, 18 and 23 days give a median of 13 days, and if the slowest parcel had taken 31 days instead of 23, the median would not have moved at all. The number a buyer remembers is the tail; the number a median reports is the middle. Both are real, and they answer different questions.
A forty-row register divided across half a dozen transport products rarely gives more than ten rows to any one of them, and often gives three or four. Three rows cannot rank anything: they can only report that three specific parcels took the times they took. What forty rows can do is describe the shape of the whole set — how wide the spread is, which weight bands cluster, whether a season added days to everything at once.
Part three: the columns that decide whether the record is usable
Most registers fail on structure rather than on volume. If the weight column holds the weight printed on a listing instead of the weight the freight bill was based on, then every cost conclusion drawn from the register is drawn from the wrong number, because postage follows chargeable weight — the greater of the packed weight and the volume converted at a divisor.
The second structural failure is a missing season marker. Holiday periods move cut-offs rather than merely slowing them, so a row from a peak month and a row from a quiet month are not two samples of one process. Without a marker, the register silently mixes two populations and the resulting spread looks like unpredictability when it is partly calendar.
The third is the published range. Recording only a midpoint discards the only information a carrier actually supplied. Recording both ends keeps the promise, its width, and the date on which the promise was read — which matters because promises change and a register with undated ranges cannot be compared with a later one.
| Column | What it can answer | What it cannot answer |
|---|---|---|
| Payment date | How long the seller took to dispatch | Whether the delay was the seller’s or the courier’s |
| Warehouse release date | When the international leg began | Whether inspection was requested in time |
| Transport product | Which product the parcel travelled on | Whether a different product would have been better |
| Chargeable weight | What the freight bill was based on | Whether packing inflated the volume |
| Published range | How wide the promise was when read | Whether the promise is a measurement |
| Arrival date | Total elapsed days for this parcel | Which stage consumed the variance |
| Season marker | Whether the row sits in a peak period | How much of the delay was seasonal |
Part four: what the catalogue side of the same workbook shows
A dispatch decision is made about goods, so the register usually lives beside a second block of rows describing what was bought. This site’s own sample of that second block is a snapshot of 218 catalogue listings taken on 2026-09-29. Its median is $36.18, its quartiles are $22.53 and $53.32, its tenth and ninetieth percentiles are $15.57 and $72.30, and its full range runs from $3.62 to $190.88. The interquartile gap, the span that holds the middle half of the sample, is $30.79.
Two features of that block matter to a transit register. The first is that most parcels are cheap relative to their postage: 44 rows sit under $20 and 56 sit between $20 and $34.99, which together are 100 of the 218 rows, so a large share of dispatches carry low merchandise value and comparatively high freight. The second is that the expensive tail is thin — 15 rows at $80 and above — so conclusions about high-value shipments rest on few observations in the catalogue just as they do in a register.
The sample is also a reminder about catalogue metadata. Of those 218 rows, 216 have distinct titles, so two titles repeat, and 52 rows carry no note text at all, which is 23.9% of the sample. Every row in the sample also shares one badge, so the badge separates nothing. None of this is transit information; it is what the price side of a workbook looks like when it is read honestly rather than treated as a complete description of the goods.
| Category lane | Rows | Median | Lowest–highest |
|---|---|---|---|
| Popular | 24 | $36.19 | $13.38–$140.80 |
| Jackets | 23 | $47.72 | $11.15–$153.17 |
| Headwear | 23 | $16.87 | $8.53–$36.18 |
| Accessories | 22 | $50.47 | $3.62–$190.88 |
| Shoes | 21 | $55.62 | $27.95–$123.93 |
| T-shirts | 18 | $26.34 | $16.69–$53.09 |
Part five: reading the record back after arrival
The value of a register is realised after delivery, when the parcel can be compared against what was declared. Contents against the declaration, condition against the intake photographs, and the verdict recorded at the time against what actually arrived: three comparisons, all of which need the register to have kept the relevant dates and references. A row that records only an arrival date cannot support any of them.
Timing matters here too. A return window closes earlier than most buyers assume, and the count usually starts from an event in the warehouse rather than from delivery at the door. A register that holds the release date makes that deadline computable; one that holds only a tracking number leaves the buyer estimating while the window runs out.
Keeping the record also changes the next decision. A line chosen by habit becomes visible as a habit when six rows show the same product chosen for parcels of very different weights. The register does not need to prove that the choice was wrong; it only needs to show that the choice was repeated without reference to the weight band it was made for.
The counterargument: more rows do not repair a biased register
The obvious objection is that forty rows are simply too few, and that a register of four hundred would settle the question. Volume helps, but it does not touch the deeper problem: a register contains the parcels that arrived. Parcels that never arrived, that were returned at the destination, that were seized, or that were written off are exactly the rows a buyer is least eager to add, and they are the rows that carry the most information.
A second bias is selection. Parcels are usually dispatched in batches with similar contents and similar weights, so a register accumulates clusters rather than independent samples. Six rows from one holiday period share a queue that none of them caused. The spreadsheet presents them as six observations, and no arithmetic performed on them can undo the fact that they are one event observed six times.
A third is the habit of recording the outcome that was recorded by somebody else. If the arrival date is copied from a tracking page that stops updating at the destination handover, the register is measuring tracking behaviour as much as transport. None of these objections makes the register worthless; they make it a record of a process with known blind spots, which is a different and more useful claim.
Conclusion and boundary conditions
Forty rows support three kinds of statement. They support descriptions of spread: this set of parcels arrived between these two points, with this much distance between the middle and the edges. They support weight and cost observations, provided the weight column holds chargeable weight and the cost column holds what was billed. And they support seasonal comparisons, provided the season is marked in the row and a peak period is not compared with a quiet one as if they were the same regime.
They do not support rankings. Claiming that one transport product is faster than another requires enough rows per product to separate the product’s behaviour from the seller, the warehouse, the season and the destination clearance, and a forty-row register rarely reaches that density. They also do not support prediction for a single parcel: a spread is a property of a set, and any individual parcel may land outside it.
The boundary conditions are therefore worth writing into the register itself: state what the rows contain, state what is missing, and state which columns are promises rather than measurements. Measured transit days for any named transport product remain unverified in this note, as do the queue lengths behind any particular gap. What the register can do without any of that is stop the same mistake being made twice — and that is the reason to keep forty rows at all.