Data studies built from public registries, then pitched

Every study we build starts from a dataset somebody else is legally obliged to publish. For French property that is DVF, the transaction register the tax administration releases on data.gouv.fr twice a year, covering every recorded sale since 2014 apart from Alsace-Moselle and Mayotte, which sit under a different land registry. A figure drawn from it can be recomputed by the journalist you send it to. A figure from a thousand-person survey panel cannot, which is why we do not run them. Where the numbers come from: DVF for transaction prices and volumes down to the parcel. The ADEME register of energy performance diagnostics, which is how you say something defensible about F and G rated housing stock. INSEE for hotel and campsite occupancy by region and month, and for household composition. Base Sirene for business openings and closures by activity code. Eurostat where a French figure needs a European frame. Where the question needs data nobody publishes, we collect it ourselves and state exactly how: observations on fixed dates across a defined set, with the collection script and the raw file published alongside the finding. Choosing the question: Before any collection, we check that a desk has quoted a comparable figure in the last eighteen months. If nobody has cited a number of that shape recently, the appetite for it is theoretical, and we will propose a different question rather than take the brief as written. This is the step most studies skip and it is the reason most studies earn nothing. What gets published: A page on your domain with the finding in the first sentence, the method under it, the limits of the data stated plainly, and the raw file downloadable as CSV under a licence permitting reuse with attribution. Charts as static images, legible at the size a news site will crop them to, with the numbers behind them in a table so a picture desk can rebuild them. No interactive map unless the data genuinely needs one, and no infographic, which has not earned a link from a serious desk in a decade. The pitch: Two to four desks get it under embargo, chosen because they cover the beat and not because of a domain rating. The embargo is real and we honour it, which is the whole reason the second study lands more easily than the first. After publication the wider list goes out, weighted towards the trade press, which links more reliably than the nationals and which your buyers actually read. What results look like: The six studies we published in 2025 earned between 19 and 74 referring domains in the first ninety days. Two others earned four each. We will show you those two on request, because the distribution is the honest picture and both failures had the same cause: the finding was true but nobody's readers were arguing about it. Not included: Sponsored distribution, wire syndication, and anything else that buys placement. Rewriting your existing research so it looks new. Studies about your own platform's internal numbers, unless you agree to publish the file, since a proprietary figure the reader cannot verify is a press release with a chart on it. Who this is not for: Anyone who needs the study live this quarter to hit a campaign date. Collection and checking take the time they take, and a study rushed past its verification step becomes a correction issued against your brand. Also not for clients who want to approve the finding before it is calculated; you approve the question and the method, never the result.