How a record gets into the ledger, and how it leaves
This page states the rules in the form they are actually applied, so a reader can check our work rather than take our word for it. Every rule below is executable against the record fields themselves; none of them relies on a judgement we cannot show.
1. How a record is selected
Records enter the ledger by extraction from the source ledger, across its full listing and each of its category views. Duplicates are removed on the marketplace item ID, so a listing appearing in two category views becomes one record rather than two.
A record is dropped when it has no images. That is the only content-based exclusion: a record with a single image is kept and marked accordingly, while a record with no image at all is removed rather than displayed with a placeholder. The current extraction holds 195 records and dropped none for that reason.
2. How verification status is assigned
Status is derived from three fields, in this order. There is no editorial override and no manual exception.
| Status | Required conditions | What it means |
|---|---|---|
| verified | image count ≥ 5, source link present, brand or category information present | The record carries enough material to be compared against another record. |
| watching | image count ≥ 5, source link present, brand not recorded by the source | The record is usable, but one field is missing and the status records that rather than guessing a brand. |
| unverified | image count < 5, or source link missing | The record is kept and remains linkable, but it is not submitted for indexing. |
- Source:
- Derived at build time from the record fields themselves.
- Sample:
- 195 records: 158 at or above the image threshold, 37 below it.
- Recorded:
- 2026-W40
- Known gap:
- Brand is copied from the source and never inferred. Where the source left it empty, the field stays empty.
A status is a statement about the record, not about the item. A verified record is one we can compare; it is not an item we have vouched for, and no page on this site says otherwise.
3. How the source image set is read
- Images are read from the record's own listing page, not from a listing preview.
- They are deduplicated on their file identifier, so the same image cannot appear twice.
- They are ordered by the sequence number in the file name, keeping the seller's own order.
- The count reported on a record is the length of that deduplicated, ordered set.
A listing page that yields no images is treated as unavailable: the record falls back to its single listing image and its count says so, rather than being credited with a set that does not exist. In the current extraction, 2806 images were read across 195 records, with a median of 14 per record and a maximum of 46.
4. How the index tier is decided
The threshold is 5 source images, and it is applied by one function that both the record pages and the sitemap call. That is deliberate: when the same rule generates the page's robots directive and the sitemap entry, it becomes impossible for a page to be excluded from the index and submitted for it at the same time.
Records below the threshold are marked noindex, follow. They remain reachable from the category tables, keep their source links, and pass no indexing budget to a page that could not support a judgement anyway.
5. How a data gap is recorded
Where a figure would need a sample this site does not hold, the field is either left to the reader as an input or the row is left out. Both choices are visible on the page rather than hidden behind a note.
| Field | Why we do not measure it | What the page does instead |
|---|---|---|
| Link availability | Would require a request to the marketplace for every record, and the result would be a snapshot that goes stale immediately. | Reads link structure only, and states on the instrument page that availability is not tested. |
| Parcel transit time | Would require a tracked sample of parcels we have not shipped. | Publishes the dispatch windows that are calendar facts, and labels them as structural rather than observed. |
| Loss probability | Would require a sample of lost parcels large enough to estimate a rate. | Leaves the probability as an input on the insurance instrument and says so in the result. |
| Measured sizing | Would require physical measurements with a stated source type and sample count. | Keeps the evidence table empty until rows meet that bar. |
| Duty or inspection outcomes | Would require outcomes for parcels we did not send. | Publishes thresholds and rates, which are documented, and no outcome rate. |
- Source:
- Our own field inventory, reviewed each week.
- Sample:
- Five fields listed; all five currently unmeasured.
- Recorded:
- 2026-W40
- Known gap:
- This list can grow. A gap that closes is recorded in the check log with its source and sample.
6. What we refuse to publish
- Discount codes we have not verified. A code list with placeholder strings and no expiry is worse than no list.
- Availability or stock status. It changes within hours, and copying it would turn a records page into a shop page.
- Ratings and success rates. Anything with a percentage attached needs a denominator, and we do not have one.
- Brand inference. Where the source left brand empty, we leave it empty and the status records it.
- Prices presented as live. Every price on this site is a recorded reference with the extraction date attached.
7. Corrections and reports
Send corrections to records@spreadsheetkakobuy.click. A useful report names the record ID or the page address, states what is wrong, and says how you know. Reports that identify a wrong figure are the most valuable; reports that disagree with an item's quality are outside what this site records.
Accepted corrections appear in the check log. Parameter changes are logged with the old value and the new value; record status changes are logged with the record they affect.
Extraction record
- Extracted:
- 2026-09-29
- Records:
- 195
- Source images:
- 2806
- Index threshold:
- 5 images