The short answer
Satellite data gives investors an independent reading of physical activity: cars in store parking lots, oil in storage tanks, ships in port, the condition of crops, the pace of construction and the heat coming off industrial plants. Because it is observed directly rather than reported, it can show a change weeks before it appears in a company's results or an official statistic.
Its value depends on three things: turning images into a consistent time series, proving that the series tracks the figure you care about, and settling with your compliance team where the data comes from and what its license allows. Treat it as evidence to weigh, never as a forecast on its own.
What satellite data adds to investment research
For hedge funds, family offices and other investors, satellite imagery is a form of alternative data: information that does not come from filings, exchanges or official statistics. Satellite analytics is the work of turning the images into measurements. Its strengths:
- Independence. The reading does not depend on what a company discloses or how it defines its own figures.
- Timing. Companies report quarterly and official statistics arrive late. Imagery can be read while the period is still under way.
- Coverage. It reaches private companies, competitors and regions where official data is thin, and works site by site.
Its limit is just as clear: satellites measure physical activity, not revenue or margins, and the link between the two has to be built and tested. The wider discipline is covered in our guide to geospatial intelligence for business.
Common uses, and what to watch for
| Use | What is measured | Caveats |
|---|---|---|
| Retail and logistics activity | Cars in store parking lots; trucks and trailers at distribution centers | Cars are not sales. Online orders, covered parking and the hour of the image all weaken the link. |
| Storage levels | How full floating-roof oil tanks are, from the shadow the wall casts on the roof; the size of coal, ore and other bulk stockpiles | Fixed-roof tanks show nothing. Volume says nothing about grade or ownership. |
| Construction pace | Cleared and graded land, foundations, cranes and new structures at plants, warehouses and housing | Visible progress is not completion. Projects pause, and interior work cannot be seen. |
| Crop condition | Planted area and plant health through the season, from how vegetation reflects red and near-infrared light | Crop type has to be confirmed on the ground. Cloud hides fields in key weeks. |
| Industrial activity read from heat | Thermal output of plants, smelters, steel mills and refineries, and gas flaring | Heat shows a plant is running, not how much it produces or at what margin. |
| Port and shipping activity | Ships at berth and at anchor, container stacks, vessel counts; radar picks out ships through cloud and at night | More ships waiting can mean congestion, not more trade. |
Each use needs the right sensor. Counting cars needs very high resolution, where a pixel covers well under a meter of ground; crops and heat can be read far more coarsely, often from free public missions. Construction progress is also the starting point for tracking a construction pipeline.
Revisit and timing
Revisit is how often a satellite, or a constellation, can image the same place, and it sets how fine a time series can be. Free public missions image most land every few days to every couple of weeks at moderate resolution: enough for crops, heat and large construction, too coarse for cars. Commercial constellations can image a site daily or more often, at much higher resolution, under a paid license. Radar sees through cloud and at night.
Two details catch people out. Many optical satellites pass over at about the same local time on each visit, so an image is a snapshot of that hour, not the day's traffic. And the capture date differs from the date the data reached you; only the second tells you what you could have known at the time.
Turning imagery into a time series
A single image is an anecdote. Investors need a series: the same measurement, taken the same way, over time.
- Fix the footprint. Draw the exact boundary of each lot, tank farm or berth, and keep it identical every time.
- Define one measurement. A car count, a fill percentage, an area under construction, a vegetation index, heat above the surrounding ground.
- Extract it the same way every time. When a sensor or source changes, record the date and compare old and new over a period of overlap.
- Grade every observation. Record cloud cover and viewing and sun angles, and drop poor readings rather than fill them with guesses.
- Aggregate and align. Roll sites up to a company, region or commodity, and cut the series to the quarter or month it is meant to anticipate.
- Keep it point in time. Store each value with its capture date and the date it became available, so later tests cannot use information that did not yet exist.
Test it against reported figures before relying on it
A satellite series earns a place in an investment process only once it has been shown to track the figure it is meant to anticipate.
- Pick the target. Same-store sales, reported inventories, official crop estimates, port throughput, a producer's output.
- Compare changes, not levels. Did the series rise and fall when the target did, and by a similar proportion?
- Build on early periods, test on later ones. Settle the method on one stretch of history, then check it on periods it never saw.
- Ask what it adds. A series that only repeats what analysts already expect is worth little.
- Retest when the business changes. If a retailer shifts to online orders or a terminal changes what it handles, the link between image and result changes too.
History is short. Five years of quarterly results gives only 20 reported figures to test against, a thin basis for strong conclusions and a good reason to prefer simple methods.
Compliance: sourcing and licensing
Review satellite data like any other alternative data before it informs a decision. The usual questions:
- Where does it come from? Which satellites and which operator, and whether the vendor has the right to license it to you.
- What does the license allow? Internal research or client reports, redistribution, and who owns what you derive from it.
- Does it contain personal information? Very high resolution imagery, and data combined with it, can fall under privacy rules in some jurisdictions.
- Could any part be confidential information? Insider trading rules apply to alternative data too. Ask whether anything in the set, or combined with it, was obtained from a company or in breach of a duty owed to one.
This is general information, not legal advice. Get your compliance team's view on each data set before it is used, keep the review on file, and repeat it whenever the source or terms change.
Pitfalls
- Seasonality. Lots fill before holidays, crops follow the season, and storage builds and draws with demand. Without a year-on-year comparison, the series mostly measures the calendar.
- Cloud cover. Optical images can be missing for weeks, and the gaps cluster in the cloudiest months, so they bias the series as well as thin it. Radar helps; filling gaps with guesses does not.
- Small samples. Hypothetical example: a chain has 1,000 stores and 50 of them are imaged clearly and often. That is 5% of its footprint, and if those 50 lean toward one region, so will the reading.
- Overfitting. With few data points and many ways to cut the imagery, it is easy to find a method that fit the past by chance. Fix the method before testing, keep it simple, and judge it on periods it was not built on.
Crowding is the quieter risk: when many investors watch the same lots and tanks, what the data reveals is priced sooner.
How Lualdi Advisors does it
The Advisory practice and MITHRIL
For hedge funds and family offices, SIGMA watches price, flow, volatility and the operational data beneath them, and tells you when the pattern breaks, before the shift has a story. Every finding carries a date, a confidence grade and reasoning an investment committee can check.
For corporations, MITHRIL reads the physical footprint of any operation on Earth, from stockpiles and truck traffic to the construction that shows next year's capacity today, and our analysts turn it into a brief a board can move on.
Its layers, among them construction progress, vehicle counts, storage levels, vegetation stress and thermal output, are cited, validated models. Every number traces back to the original pass; where that record cannot be produced, the value is not drawn. MITHRIL reads optical, radar, thermal and hyperspectral imagery, sees through cloud and dark, and revisits the ground that matters in hours, not weeks.
Questions
What is satellite data used for in investing?
Investors use it to read physical activity directly: cars at stores, trucks at warehouses, oil in tanks, ships in port, crop condition, construction progress and heat from industrial plants. It is one of the best known forms of alternative data.
How often can a satellite image the same place?
Free public missions revisit every few days to every couple of weeks at moderate resolution, while commercial constellations can image a site daily or more. Cloud can block optical images for long stretches; radar sees through it.
How do you know if a satellite data set is useful?
Test it against the reported figure it is meant to anticipate, building the method on early periods and judging it on later ones. Keep only series that track changes in the target and add something analysts do not already expect.
Does satellite data need compliance approval?
Most firms review alternative data before using it: its source, its license, any personal information and whether any part could be confidential information. Get your compliance team's view on each data set before it informs a decision.
What are the main pitfalls of satellite data for investing?
Seasonality that looks like real change, cloud cover that leaves gaps, small samples that misrepresent a whole company, and methods overfitted to a short history.
