Your last collection page is thinner than your first

We measured this on our own site this week rather than guessing at it. Nitaris publishes seven pages that each explain the product for one kind of seller. They were written over several months, all to the same brief, and they should be interchangeable in weight. Counting the words each one actually renders: 1,721, 1,706, 1,565, 1,543, 1,294, 1,281 and 1,259. The thinnest is 27% lighter than the heaviest, and until someone put the seven numbers in a column, nobody on this team could have told you which was which.

The first one is written, the rest are adapted

The first page in a set gets made out of nothing. You decide what the section is for, you find the examples, you cut the paragraph that was not earning its place, and you end up with something that was genuinely thought about. The second page starts from a copy of the first, which is the sensible thing to do — it is also the moment the set stops growing and starts eroding.

Adapting goes one direction. A sentence that named the first category has to come out of the copy, and the specific replacement for it takes real work, so quite often it does not get written. A list of six examples becomes four, because only four came to mind that afternoon. None of these are mistakes anyone would catch in review, because at the moment you review page six you are reading page six, and it reads fine.

A floor is not the same as a comparison

Whatever quality bar you keep, it is almost certainly a floor applied to one page: enough words, an image, a heading, no empty sections. Our own automated check holds every page on this site to a 300-word minimum, and all seven of these pages clear it more than four times over, so the check has nothing to report and is right not to report anything. The gap is not between a page and the standard. It is between a page and its siblings — and that is a comparison, not a threshold, which is why the same thing happens to shops that are careful about consistency of look.

This is the uncomfortable part: a consistent template makes the problem harder to see, not easier. When all seven pages carry the same layout, the same heading structure and the same photography treatment, the thin one is camouflaged by everything it does have. You scan it, you recognise the shape, and the shape is correct.

Sort the set once

The fix is embarrassingly cheap, and it is a sort rather than an audit. Take every page in one set — your collections, your category pages, your per-material or per-occasion landing pages — and get a word count for each into a single list. Order it. You are not looking for a number to hit; you are looking at the bottom three, which are almost always the three you made last, and asking what each of them is missing that the top of the list has.

• Examples. The top page usually names five or six products; the bottom ones name two.
• The objection. The first page you wrote answered the thing customers actually ask about that category. The adapted ones inherited the answer to a different category's objection, or dropped it.
• Sizes, materials, care, fit — the concrete detail that only applies to this group and therefore could not survive being copied from another one.
• Media. A set that was adapted tends to reuse the first group's photography far longer than anyone intended.

Then fix one. Not all of them, and not today — the bottom entry on that list is where an hour buys the most, and the next time you sort, it will be a different page at the bottom. The same logic applies to the moment a set first appears: the second thing you add is the one that decides whether you have a set at all.

None of this is an argument for padding. A page that says what it needs to in 900 words is finished at 900 words, and a 1,700-word page can be the worse one. The sort is a question, not a verdict: it tells you where to look, and it is the only view of your own site that shows you the page you have stopped reading.