Satellite Imagery Signals
Satellite images of parking lots, oil storage tanks, and shipping ports let analysts literally count physical economic activity from orbit — an unusually direct alternative to waiting for a company to report it.
Prerequisites: Sourcing and Vetting Alternative Data
Commercial satellites now photograph most of the earth's surface repeatedly, at resolution good enough to count individual cars in a parking lot or measure the shadow cast by the floating lid of an oil storage tank. Satellite imagery signals turn these repeated photographs into a data feed: a computer vision model scans the images and counts or measures something economically meaningful, tracked over time.
Satellite imagery lets an analyst directly observe physical activity — cars in a lot, oil in a tank, ships at a port — rather than waiting for a company or government agency to report it, using computer vision models to convert raw images into a countable, trackable number.
What gets measured
Retail parking lot counts are used as a proxy for store traffic, much like phone-based geolocation data but derived from cars rather than devices. In commodities, the shadow cast by the floating roof of an oil storage tank reveals how full the tank is, letting analysts estimate crude oil inventories at storage hubs well before official government inventory reports. Agricultural applications measure crop health across farmland using the color and infrared signature of vegetation, feeding into crop yield forecasts ahead of harvest.
Worked example
An analyst tracking a major oil storage hub uses satellite images taken every few days to estimate that floating-roof tank levels across the hub have risen 8% over the past month, implying inventories are building faster than the market expects. That estimate, produced from image analysis, arrives well before the next official government inventory report, and if it turns out to be right, gives the analyst an early read on a supply glut that should eventually pressure prices.
What this means in practice
The accuracy of any satellite signal depends entirely on the underlying computer vision model correctly identifying and measuring what it is looking for, and models trained on one region's or one company's facilities do not always generalize well elsewhere — a parking lot count model tuned for suburban US stores may miscount at a densely packed urban location with different lighting and layout. Cloud cover, satellite revisit frequency, and image resolution also limit how often and how reliably any single location can actually be observed, so gaps in coverage are common and need to be accounted for rather than silently treated as "no change."
Ask a satellite data vendor for their model's validated accuracy against known ground-truth numbers (like a company's own reported store counts or a government's inventory report) before trusting the signal — a compelling image is not the same as a validated measurement.
Further reading
- Orbital Insight and RS Metrics methodology notes on satellite-derived economic indicators