Data Studies Built From Operational Records You Already Hold

The studies that get cited in logistics and energy are built from records the company already keeps and considers dull: dispatch timestamps, warranty claims by failure mode, meter readings, tender outcomes, yard dwell times, fleet telematics. We do not commission survey panels. Five hundred bought responses produce a chart and no citations, because the desks in these sectors read the methodology before they write the headline, and where they do not, the analysts their readers trust do it for them. Where the data comes from: Two working days with operations, quality and IT, and not with marketing, who will suggest a survey. We are looking for a series you hold that nobody outside your company holds, long enough to show a trend and granular enough to break out by region, sector or component. A ten-year warranty table split by failure mode is a better starting point than any dataset available for purchase, and it is usually sitting in a system somebody built in 2011. The claim test: Before anything is built we write one sentence: falsifiable, containing a number, publishable as a headline with nothing added to it. If that sentence cannot be written from your data, the engagement stops there and you pay for the work done rather than for a campaign. About one dataset in three fails this test. A study that proceeds without passing it becomes a PDF that gets three links, two of them from your own suppliers. Clearance before construction: Aggregation thresholds set so no single customer's volumes can be read back out of the published cut. Sign-off from commercial on what the figure tells a competitor who already knows your capacity. Where the data touches counterparties, the contractual position on publishing it, in writing, before a line of the asset is built. This step has killed studies at the last minute for other agencies; we take it first, when killing one is cheap. The public spine: Your figure needs something to sit against. Eurostat, ISTAT, Terna's load and dispatch data, GME price series, port authority throughput reports, ACEA registrations. The story is rarely that your data says X. It is that your data says X while the public series says Y, and the gap is the part that gets quoted and the part that gets argued with, which is the same thing as being linked. The asset: An HTML page, not a gated PDF and not a deck. Methodology in full: sample definition, period, exclusions, and a stated account of what the data cannot show. A downloadable CSV of the aggregate table, because the people most likely to cite you will want to check it. Charts with axis labels inside the image so they survive being screenshotted into somebody else's article, and a reuse licence stated on the page. We do not: Put the study behind a form. Publish it under an executive who did not do the analysis. Write the industry-report version, where a copywriter assembles other people's public figures into a branded document; that is not a study and it earns nothing. Guarantee pickup, either. The figure interests a desk or it does not, and we tell you which we expect before you commit the budget rather than afterwards. Wrong fit: Companies whose data sits in a system nobody can export from without a vendor project and a change request. Companies whose legal function will not clear one public figure. And anyone who wants the study to be about their product. The study is about the sector; your name appears on the byline and in the methodology, and that placement is the entire mechanism by which it works.