Original data studies built to be cited, not shared

Roughly one study idea in three dies before anything is built, usually because someone has already published the answer or because the data cannot carry a claim strong enough to be worth a journalist's time. That is the useful part of this work. A study earns citations when the finding contradicts something the trade currently believes; a study that confirms what everyone assumed gets read and forgotten. The question comes first: We start from a disagreement, not from a keyword. Does the industry-standard size chart still describe the bodies it claims to? What does a garment actually sell for on Sellpy, Tise or Vestiaire two years after it leaves the site, and which brands hold value? How many beauty products carrying a specific claim list an INCI that contradicts it? Before a line of collection code is written I check that the answer is not already public, and I write down the finding that would make the study worth publishing and the finding that would make it worthless. If the second one is what comes back, it gets published as that, or shelved. What gets built: A collection method with a stated window, stated sample and stated exclusions. Product listing data captured weekly and held for at least 26 weeks before any seasonality claim is made, because a shorter window describes a promotion, not a pattern. Size chart measurements read from the brands' own published tables, so the source is checkable. Confidence intervals shown, not implied. A three-point gap is not written up as a trend. Publishing so it can be cited: One permanent URL that never moves into a blog category and never redirects. Static charts with the underlying numbers in an HTML table directly beneath them, because interactive dashboards get shared and static numbers get quoted. Dataset published as CSV under CC BY 4.0 with a DOI minted through Zenodo, so reuse carries an attribution obligation and academic citations become traceable. Schema.org Dataset markup on the page. A methodology section that states plainly what the data cannot show, which is the section journalists read to decide whether you are serious. What I will not do: No panel survey of a thousand consumers written backwards from a headline somebody already agreed on. No "X per cent of shoppers say" research, which is an opinion poll wearing a lab coat. No infographic built as outreach bait. No city rankings, no "best places for" indexes, no studies whose only purpose is that a local paper will mention the brand name. These land links and they cost you the ability to publish anything anyone believes afterwards. Refresh and afterlife: An index is worth more in year two than year one, because the second edition creates a comparison and the comparison is the story. Where a study is built as a repeatable index, you get the collection scripts, the raw data and a written protocol so your team or I can run it again without reconstructing the method from the article. Not for: Brands who want the study to promote a product. The moment a finding flatters the sponsor, a trade desk stops treating it as data. Brands who need the result to say a specific thing. Brands who cannot wait ten weeks. And brands who want this and the outreach as one line item at a small price; building the asset and pitching it are separate pieces of work priced separately, and the second one only makes sense if the first one held up.