The 2026 Kakobuy Spreadsheet Year in Numbers

One date, 2026-09-29 in week 40 of the year, is the entire time span of this record. That single extraction is the honest shape of the ledger's first year: one snapshot, 195 entries, and no second observation of the same fields to compare it against. Nothing in this article is a growth figure, a rate, or a year-over-year change, because a change needs two measurements and this record has one.

What the year does contain is a complete census of one pool on one day. Every entry comes from the eight category views of a single source marketplace, which makes the pool narrow in a way that is easy to forget when reading a category table. All 195 records are attributed to Weidian; no other platform contributes a row. The narrowness is a limitation of the record, and it is also what makes the figures reproducible: one source, one date, one set of extraction rules.

The category table below is the centre of the year, and it is the only place where the pool divides into groups large enough to compare without the comparison collapsing into single rows. Category counts run from 24 to 26 entries, which is close enough to even that no category dominates the pool by volume. What differs is price.

Two features of that table are worth naming. The first is the spread between category medians, which runs from $16.22 in Headwear to $68.68 in Shoes, a factor of roughly four. The second is the Accessories row, whose median of $26.77 sits above a maximum of $451.10: the cheapest entry in the whole pool and the most expensive one are both in that category, which is the clearest argument on this site against reading a category by its median alone.

A same-basis comparison with 2025 cannot be made from this data, and the reason is structural rather than technical. No 2025 extraction of these fields exists on this site. There is no earlier price for any entry, no earlier image count, no earlier record of which entries were listed, and no earlier version of the parameter table. Producing a percentage change from a single observation is not a difficult calculation; it is an invented one, and it would be indistinguishable on the page from a real figure.

What can be compared is a different kind of number: the magnitude claims that circulate in public discussion of agent spreadsheets, where figures such as 2,200 or more entries, 6,000 or more, 10,000 or more, and 100,000 or more appear. Those claims differ from each other by a factor of roughly 45 between the smallest and the largest, which is far more than any plausible difference in what a spreadsheet contains. The spread is explained by the absence of a stated basis: a count without a date, without a platform mix, and without a deduplication rule can be almost any number and remain defensible to the person who wrote it.

For next year to support a real comparison, five things would have to stay fixed while a second extraction is taken: the date, the set of source views, the field definitions, the deduplication rule, and the index threshold. Change any one of them and the two observations measure different things, which is the ordinary failure mode of year-over-year tables. The requirements are set out in the last section because they are the useful part of this section, not the apology.

Three changes in the wider picture shaped what a record like this has to carry during 2026, and each one shows up in the field structure rather than in the price figures. None of them is a change this site measured; they are changes in the environment the record describes.

1. European discussion of assessing duty by category rather than by item. Where a category becomes the unit of assessment, three items of the same kind can count as one unit instead of three, which changes how a basket is described before any value question is asked.
2. Public discussion of threshold adjustments across several destinations, which is why the six-country parameter table on this site carries a check week on every row rather than a single publication date.
3. The arrival of row-level check dates as ordinary practice. A row that states the week it was checked can be used by somebody who did not write it; a row without a date cannot, and the difference costs nothing to produce.

The third change is the one this ledger can demonstrate from its own structure, because every one of the 195 rows carries the review week 2026-W40. That field did not exist in the shared sheets this site was built to replace, and its absence is the single most common reason a reader cannot tell whether a figure is current. Adding it converts an undated claim into a dated one without changing the claim itself.

A useful annual article needs two observations and a stable basis. This one has one observation, so the honest version of a forward-looking section is a list of requirements rather than a set of forecasts. If the requirements are met, the next edition of this article can contain a real comparison; if they are not, it will contain another census, which is still worth publishing and is a different document.

• The image threshold will keep binding, because it is the one field comparison that can be applied to every row without a request. A future pool with thinner galleries will produce a larger unverified count for that reason alone, which is a property of the rule rather than of the items.
• The platform mix is the largest unknown. Every row here comes from one marketplace, and a second platform would change the price distribution and the image distribution at the same time, which is why it is recorded as a field rather than described in prose.
• Category-level assessment questions will keep arriving, because the difference between one unit and three changes a total without changing a basket, and no amount of value planning resolves it.
• The most likely improvement to this record is not a bigger pool but a second date on the existing one. One dated pool can answer questions about spread; two dated pools can answer questions about movement, and the second is what readers actually ask for.

The full record sits at /vault/, with the eight category views gathered beneath it, and each figure in this article can be traced back to the rows that produce it. That trace is the only claim this year can make with confidence: everything counted here was counted once, on a stated date, by a stated rule, from a stated source.