CNFans Sheet Path
← All notesReviewed

Rebuilding a Table Row From a Product Page, Field by Field

1,471 words · about 7 minutes · note 35 of 40 · Step 2

A link table is only as good as the rows inside it, and most bad rows are not invented — they are rebuilt by hand from a product page and quietly lose a field along the way. The rebuild is mechanical once the order of operations is fixed, because every field on the page maps to exactly one slot in the row. What follows takes the seven slots in the sequence that prevents rework, states what a correct entry looks like at each step, and names the mistake that actually happens there. The worked numbers come from this site’s 218-listing snapshot taken on 2026-09-29.

What has to be open before the first field

Five things belong on screen before any copying starts: the product page itself, the order form you will paste into, the payment method you will actually use, the delivery address record in the form the carrier will read, and somewhere to hold the finished row so it can be reviewed as a block rather than field by field.

The address record deserves the emphasis. A row can be flawless and still fail if the address it connects to lacks a mobile number, spells a street differently from the identity document, or omits a postcode segment. Checking that once at the start is cheaper than discovering it after a parcel is in transit.

One rule governs the whole rebuild: copy, do not summarise. Every field below is a transcription task, and the moment a field becomes a paraphrase it stops being checkable against the page it came from.

The sequence matters as much as the copying. Prices are easier to read once the option string is settled, because a page with forty styles usually quotes one price range for all of them; weights are easier to defend once the category is known; and the ratio check at step five only works if the two currency figures were transcribed before anything was rounded. Rebuilding in a different order produces the same row on a good day and a row with one silent error on a bad one.

Step 1: separate the two identifiers in the address

Operation: read the product address and split it into a descriptive slug and a numeric identifier. A typical address takes the form of a title-derived slug, a numeric id, and a short marker suffix; the slug describes the item, and the id is what a service matches on when two titles look alike.

Expected result: two tokens stored in separate fields, with the id copied complete and without spaces. Common error: pasting the whole address into the id field, or trimming the id to its leading digits because the remainder looks like noise.

Step 2: the title, copied rather than shortened

Operation: copy the item title exactly as the page writes it, including any bracketed or parenthesised count such as a style count or a set size. The bracket is not decoration; it is the part of the title that tells you how many variants sit behind one price.

Expected result: a title string that matches the page character for character. Common error: tidying the title — dropping the bracket, expanding an abbreviation, or fixing what looks like a typo. A cleaned title is no longer the seller’s title, and a mismatch between the row and the page is the first thing a reviewer notices.

Step 3: the option string

Operation: copy the option wording the page uses for colour, style or size, keeping its numbering intact when the seller numbers styles rather than naming them. Where the page lists sizes as a sequence of letters, transcribe the sequence rather than choosing from it, because the choice belongs to the order and not to the row.

Expected result: an option string that a second person could match against the page without asking a question. Common error: translating a numbered style into a description — writing the black one when the page says style 3. The buyer reading the row is matching characters, not intentions.

Two structural facts about option strings are worth holding onto. The first is that a style count and a size list are different objects: a page offering forty styles carries forty variants at one price, while a page listing six sizes carries six variants inside whichever style you chose. The second is that the row only needs the part you intend to order, quoted in the seller’s own words — not the full variant list, which belongs to the page rather than to your row.

Step 4: the price basis and the yuan figure

Operation: establish what the price is per — item, pair, set of three, or bundle — and record that basis on the row before recording the number. Then copy the yuan figure exactly as displayed, including the decimals, and note whether it is the lower bound of a range or a single price.

Correct entry: a basis field that cannot be misread plus a numeric field that matches the page. Common error: lifting the lower bound of a range as though it were the price, which produces a row that is understated by an unknown amount and an invoice that does not reconcile.

Step 5: the dollar figure and the ratio it implies

Operation: copy the dollar figure that sits beside the yuan figure, then divide the two to see what rate the pair implies. The division takes a few seconds and catches transcription slips immediately, because a correctly copied pair lands inside a narrow band while a mistyped digit does not.

Result to expect: a ratio consistent with the rest of the catalogue. The six rows sampled below imply between 6.327 and 6.336 CNY per USD, which is a tighter spread than any manual copying error would produce. Common error: assuming one rate applies throughout. Counting this site’s 218-listing snapshot row by row, 143 pairs imply about 6.33 while 75 imply about 6.05, and the mix is what produces the catalogue-level mean of 6.232 that the currency tool loads as its default. Why two bases appear in one catalogue is not established here and remains unverified.

Six rows from this site’s 218-listing snapshot (2026-09-29), with the CNY/USD ratio each pair implies
Item titleCNYUSDImplied CNY/USD
Ralph Lauren Polo Bear Sweatshirts [40 styles]223.0835.256.33
Burberry Check Sleeve T-Shirt for Men - White210.0033.186.33
AMIRI Hoodies [Multiple Styles]262.6841.516.33
Stone Island Pants and Shorts (3 Styles)237.6037.506.34
Stone Island t-shirt105.6016.696.33
Represent T-Shirt tee (35+)99.3615.706.33

Step 6: the weight you can defend

Operation: attach a planning weight to the row rather than leaving the field empty. This site’s register holds per-category templates for exactly this purpose: 900 g for a boxed pair of shoes, 550 g for a hoodie or jacket, 500 g for trousers, 250 g for a cap, belt or small bag, and 600 g for a mixed basket.

Checkpoint: a weight figure whose origin you can state — template, measurement, or the service’s own intake note. Common error: reusing the shipping weight printed on the product page, which describes the seller’s packaging rather than your consolidated parcel and can differ from the measured intake weight by enough to move a line choice.

Step 7: the note, badge and lane fields

Operation: transcribe the short description, the marker label and the lane the row was collected from, then read them once to confirm they describe the item you are holding. None of these three fields prices anything, and that is precisely why they are dropped first.

End state: a row whose provenance can be reconstructed later. The snapshot shows why these fields carry less weight than they appear to: 52 of its 218 rows carry no description at all, and every one of the 218 carries the same marker label, so the label cannot separate one row from another.

Common error: treating a marker label as a quality signal. A label applied to every row in a catalogue is a formatting convention, not a judgement about the item.

Where a description is missing, leave the field empty rather than composing one. A blank cell renders as a dash and stays honest; an invented description reads like source information and is indistinguishable from it later, which is how a row acquires a fact that no page ever contained.

Completion check

Five checks close the rebuild. Count the rows: the snapshot holds 218 rows across 216 distinct titles, so a title-based de-duplication silently loses two rows. Re-derive one ratio per row and confirm it lands near 6.33 or near 6.05, never in between. Confirm every row carries a category and a planning weight, and that the option string still matches the page. Run the finished block through the export tool once, because a row that survives a round trip through plain text is a row whose quoting and separators are sound.

Then date the block. A row is a snapshot of a page that keeps changing, so the useful artifact is the row plus the date it was read plus the rate it implies. At a snapshot median of $36.18, a four-row basket sits near $145, and a single mistyped digit in a dollar field moves that by more than a rounding error. This note was last checked on 2026-09-29.