We published our Q2 refinery estimates on July 8. Earnings season just ended. Here's how we did. Using Kpler's global commodity flow data — vessel tracking, refinery throughput, #crude# imports — we modeled Q2 production, runs, revenue, and gross margins for 20 publicly traded refiners before a single company had reported. The results: ◾ Crude #oil# runs: accurate within 2.45% of actuals ◾ Refinery revenue: accurate within 5% of actuals ◾ 93% directional accuracy on revenue YoY signals (13 of 14 tickers) These numbers were locked in three weeks before earnings. If your thesis on a refiner depends on getting runs, utilisation, and margin direction right before the print — this is the data infrastructure that makes that possible. Read more: