Original Data Studies for Manufacturing, Freight and Power

We published four data studies last year. Two passed 100 referring domains inside a quarter, one reached 31, and one reached 9. Before you commission anything we tell you which of those outcomes a given dataset is likely to produce, because the difference is almost never the design work. It is whether a beat reporter already wants to write the story and has no numbers for it. What we build studies from: Data you already own — telematics, dispatch records, SCADA histories, warranty claims, quote-to-order data — or public federal files nobody has bothered to join. EIA Forms 923 and 860 for plant-level generation and capacity. The EPA greenhouse gas reporting programme at facility level. FMCSA SAFER and the MCMIS crash file. BTS Transborder Freight and the Freight Analysis Framework. PHMSA pipeline incident records. FERC Form 1. OSHA establishment-level inspection data. Census M3 shipments and the Annual Survey of Manufactures. The work that earns citations is almost always the join: inspection history against facility emissions, crash rates against carrier size and driver tenure, plant retirements against county employment. How the engagement runs: We agree the question first, and we agree in writing that the finding publishes whichever way the number falls. Then acquisition and cleaning, with the vintage and download date of every file recorded. Then analysis. Then an internal challenge, where a second analyst is paid to break the finding before a reporter tries. Then a methodology page carrying the exact query, the caveats and the known limits of each source, plus the cleaned dataset as CSV and Parquet, ungated. Charts ship as SVG with the underlying table beside them, so a reporter can lift a number without emailing anyone for it. Distribution: Five to eight named reporters get the study under embargo, five business days ahead, chosen because they have covered this specific question before without numbers. We show you their clips before we send. After the embargo lifts: the trade wires that matter in the sector, the relevant association newsletters, and the forum and subreddit threads where operators argue about this in public. We do not buy syndication and we do not send to a purchased list. What we refuse: A survey of 200 respondents dressed up as research. An infographic that restates a press release. A study whose conclusion was decided at the kickoff. And any engagement where the client keeps a veto over an unflattering result. That condition ends the conversation at proposal stage, and we would rather lose the work there than publish something a reporter can take apart in an afternoon. What you receive: The study page and the methodology page as production-ready HTML. The cleaned dataset in both formats. The chart set. The embargo list with clip evidence for every name on it. The correspondence log. A link report at 30 and 90 days showing referring domains with anchor and destination. Not included: Primary research needing fieldwork or a licensed survey panel. Paid amplification. Ongoing dataset maintenance — if the study should refresh annually, that is its own engagement and we will say so rather than quietly rolling it into a retainer. Who this is not for: Anyone who needs this live inside a quarter for a campaign already booked. Acquisition and cleaning on federal files runs three to five weeks before anybody looks at a chart, and compressing that step is where the errors that get you corrected in public come from. Also not for companies who cannot get a data-sharing sign-off from counsel, where the study is built on their own operational records.