CSV batch thinning checker

Each row reads fine on its own. The problem is that across the export, the same opening sentence appears ninety times with a different product name in the middle, and nobody notices because nobody reads four hundred rows in order.

Every row is reduced to its overlapping four-word sequences and compared pairwise with the rows before it. Two rows sharing 80% or more of those sequences are reported as a near-duplicate pair, with the percentage and both row numbers, so the fix is a rewrite rather than a re-read.

The second view counts five-word phrases that turn up in two or more different rows. A handful is normal boilerplate; the same phrase in forty rows is a cohort that will read as programmatic to anyone, and to a helpful-content review.

Thin content here means a word count below the floor you set, and nothing more. A 155-character meta description is thin by any product-description floor and perfectly correct for its slot, which is why the threshold belongs to the column rather than to the tool.

What none of this decides is whether the copy is true, whether the pages compete with each other for the same query, or whether the catalogue should exist at all. Those need the console and, for a real catalogue, a person.