The US weekly petroleum status report is the most reliably market-moving scheduled data release in oil. It reports crude and product inventories, refinery runs, production, imports and exports for the week ending the previous Friday, and it is published on Wednesday morning.
It matters because it is the closest thing to a measured read on the balance. Production and consumption are estimated; a stock change is counted. So the weekly print is the market's most direct observation of whether supply exceeded demand in the recent past, for the one large region where that is measured weekly.
It is a survey, not a census, and the mechanics matter. Operators report to a statutory survey with a legal obligation to respond, which makes the response rate high, but it is still a sample-and-estimate process with imputation for non-response. See survey mechanics. It is also revised, so the first print is a draft. See revisions.
The market convention of trading the surprise, meaning the difference between the print and the expected figure, has a specific data problem attached: the expectation is a consensus of analyst estimates, and a genuine consensus is a commercial product. That is why petral has no consensus-versus-actual panel. The actuals are free and the expectations are not, and a surprise computed against an invented consensus would be a fabricated number wearing the clothes of a measured one.