Where Kakobuy Spreadsheets Are Heading: Drivers and Timelines

A working Kakobuy sheet in 2026-W40 looks stable, and the stability is misleading. The current pool holds 195 entries across 8 categories and 20 brands, all sourced from Weidian, carrying 2806 listing images at a median of 14 per row against an index gate of five. Of those rows, 158 clear the gate and 37 do not, and none sit in the intermediate watching state. Those four numbers describe a sheet whose shape has settled without its contents having settled at all.

What has stabilised is the scaffolding. The fee block has six rows: item subtotal, domestic shipping at $1.40 per item, commission at 5 percent, surcharge at 3 percent, freight at $45 and destination tax above a threshold. The destination layer has six rows, and the divisor set offers 5000, 6000 or 7000. Those structures have been consistent long enough to be treated as a specification rather than as a preference, which is what allows a sheet written this week to be compared with one written a month ago.

What has not stabilised is the confidence behind those rows. The destination table carries 264 checked rows for Germany, 196 for the United States, 138 for Canada, 39 for Australia, 35 for the United Kingdom and none for Poland. The gap between 264 and zero is not a gap in the threshold value, since Germany and Poland share a €150 line. It is a gap in how much has been verified behind each line, and that gap moves whenever somebody checks another destination.

The annual read at /field-notes/spreadsheet-year-in-numbers/ is where the pool-level figures are collected, and the direction of travel in those figures is the subject of the rest of this note. Three drivers account for most of the movement, and each of them pushes the sheet in a different direction.

The first driver is verification pressure. The five-image gate is a cheap rule that sorts a list into a documented part and an undocumented part, and it currently splits the pool 158 to 37, or 81 percent to 19 percent. The pressure comes from the fact that the gate is permanent: rows do not move above it on their own, and a list that grows without re-checking will see its verified share fall while its row count rises. That is a coverage problem rather than a size problem.

The second driver is destination fragmentation. Six destinations now carry six combinations of threshold, rate and handling fee, and the combinations do not scale together: Canada applies 13 percent and a CAD 9.95 fee from a CAD 20 threshold, which the first item crosses, while the United States applies nothing until $800. On top of that, the per-category duty rule means a parcel can be assessed on its composition rather than on its total, which makes the mix of categories inside a box a first-order variable in a way it was not before.

The third driver is freight and volume economics. The divisor set of 5000, 6000 and 7000 turns a single carton into three different billed weights: a 40 by 30 by 20 centimetre box bills at 4.80 kg, 4.00 kg or 3.43 kg depending on which divisor the quote uses, against an actual weight of 2.5 kg. At a marginal rate of $9 per kg, moving from a 6000 default to a 5000 divisor adds $7.20 to that one carton, and moving to 7000 removes $5.13.

• Verification pressure acts on the row layer: gate coverage of 81 percent, 37 rows below the line, and a median of 14 images per row that makes the gate cheap to apply and expensive to ignore.
• Destination fragmentation acts on the tax layer: six thresholds, six rate and fee combinations, and checked-row counts that differ by a factor of more than seven between Germany and the United Kingdom.
• Freight economics acts on the parcel layer: three divisors in circulation, a $9 per kg marginal rate, and a $12.33 spread between the highest and lowest divisor on a single carton.

The first path is the verification-first sheet, in which coverage becomes the headline figure rather than the row count. That sheet carries a gate column, a coverage percentage and a queue of rows below the line, and its totals are quoted against the documented subset rather than against the whole list. The change is small in structure and large in effect: a 195-row sheet with 158 documented rows stops being described as a 195-row sheet, and the 37 open rows become a work item with a visible size.

The second path is the destination-split sheet, in which the six destinations become six sheets or six column groups, each carrying its own threshold, rate, fee and checked-row count. The driver behind this path is the per-category duty rule, which makes a parcel composition question rather than a total question. A split sheet answers that directly, because each destination column can be assessed against its own categories, and it removes the failure described at /field-notes/why-spreadsheets-rot/, where a workbook keeps answering the question it was built for after the question has changed.

The third path is the instrument-linked sheet, in which the arithmetic leaves the spreadsheet and the sheet stores outputs. Landed cost, parcel merge decisions and link verdicts already exist as instruments, and each of them produces a field that can be pasted back into a row with a date attached. The advantage is that a formula cannot be out of date while an instrument can be re-run. The cost is that a sheet of stored outputs is only as good as the dates next to them, which pushes the discipline of dating every field from good practice into a requirement.

The three paths are not exclusive, and the order in which they arrive is likely to be the order above. Verification is a column and can be added to any sheet today. Destination splitting depends on how many destinations a reader actually buys for, which is a fact about the reader rather than about the market. Instrument linkage depends on how often the upstream instruments change their own settings, and that is a slower variable than either of the other two.

The working review cycle for everything in this note is 2026-W40, and the record of what moved between cycles sits at /vault/changelog/. The comparison note at /field-notes/practice-changed-2025-2026/ covers the changes that have already landed, so the indicators below are the ones that would move the practice further rather than restate where it stands.

The Polish row is the single clearest trigger on the list. Germany and Poland share a €150 threshold, so the only difference between a usable row and a placeholder is the handling fee and the checked-row count behind it. The moment a fee is confirmed, a line that currently says not confirmed becomes a number, and the destination block gains a seventh comparable row instead of a sixth plus an exception.

Coverage is the second indicator and the easiest to move, because it is a count rather than an external fact. The gap between 158 documented rows and 195 total rows is 37 rows, and each one closes by adding images rather than by waiting for anything. A coverage figure that stays at 81 percent across two review cycles while the pool grows would be the first sign that the gate has become a label rather than a filter.

The divisor default is the third indicator and the most expensive one to miss, because it changes the freight line on every row at once. The spread between a 5000 and a 7000 divisor is $12.33 on a single carton, which is close to the headwear median of $16.22 and larger than the $8.37 that a 7000 divisor costs on that same box. A sheet that records which divisor was used can be re-priced when the default moves. A sheet that records only a freight figure cannot.

Taken together, the three drivers and the seven indicators describe a workbook that will keep its shape and keep changing its numbers. The scaffolding of six fee rows, six destinations and three divisors is unlikely to move soon. The readings inside it, from 81 percent coverage to a Polish fee that does not yet exist, are the parts worth re-checking at every cycle, and the place to start is the changelog entry for the week being reviewed.