How a product gets into the catalogue
The filter decides one thing about every item: whether it belongs in the catalogue at all. This page describes what that decision is based on, and what it deliberately refuses to do.
The problem this filter is for
On a large marketplace the same garment appears under dozens of sellers. Same photographs, same description, often the same factory, and prices that range from three dollars to twenty. The reviews are not necessarily written by people who bought the item, and what the fabric actually is cannot be read off the listing.
This is not a lottery. An experienced buyer who knows what to look for does find good products there, by reading composition tables, comparing a seller's history against the item's review pattern, and discarding anything that does not add up. The method works. What it costs is an evening, and what it runs into is a ceiling: a person can hold a few dozen listings in their head before the comparisons blur. An automated scan works at a different order of magnitude, and applies the same standard to the last listing as to the first.
That gap between what a careful person can check and what a machine can check is what this site is built on.
Two layers of evaluation
The supplier-dependent layer asks whether the shop behind the listing can be relied on at all. It looks at how long the shop has been trading, where its rating places it among comparable shops, whether its stated delivery times match what buyers report, and whether its reviews behave like reviews from real buyers rather than a block of text posted in one week.
The category-dependent layer asks what makes an item good in its own category, and its criteria differ from one category to the next. In kids' clothing that includes the fabric composition, whether sizing matches what the listing states, and how the garment holds up to repeated washing. Adding a category means defining what quality means in it; the list is not fixed, and it also varies with what a given supplier discloses.
An item is evaluated on both layers. Failing either one keeps it out, which is why the rejected page includes items from shops that are perfectly reputable.
The gate is pass or fail, and it is the same for everyone
An item either passes the filter or it does not. There is no partial pass, and no setting anywhere on this site that lets a visitor loosen the standard. That is a deliberate constraint: if the threshold could be adjusted per person, then "passed the filter" would mean whatever each visitor had set it to, which is to say nothing at all.
Personal preferences do have a place, but only after the gate. Filtering by age, by season, by item type or by price changes the order and the subset you are shown, drawn entirely from what already passed. A preference can never pull in an item the filter rejected.
Why the thresholds are relative, not absolute
A criterion is not phrased as "rating above 4.7" but as a position within the category, such as the top fifth of shops selling in it. There are three reasons for that.
The first is catalogue size. An absolute cutoff yields however many items happen to clear it, which may be four hundred in one category and forty in another. A relative threshold yields a predictable share of what was scanned, which is what a catalogue can actually be built on.
The second is that rating scales are not comparable across suppliers. A marketplace where almost every shop sits above 4.6 and one where the same quality reads as 4.1 would need different absolute numbers to mean the same thing. A position within the category survives that difference.
The third is portability. Opening a new category with an absolute number means guessing at what good looks like there before having seen it. A relative threshold defines itself against whatever is in that category, and starts working on the first scan.
Why rejected products are published
A catalogue that shows only what passed asks to be trusted without evidence. Anyone can present a list of products and describe it as filtered.
The rejected page carries items that were evaluated and kept out, each with the criterion it failed and what was found. It is the part of the site that can be checked: read a rejection, go look at the listing, and decide whether the call was right. That check is only available if the rejections are published.