Disclaimer: This article is for educational purposes only and isn't financial advice. Trading involves risk, including loss of principal. Talk to a licensed financial advisor before making investment decisions.

The Edge Isn't in the News Anymore

For most of Wall Street's history, beating the market meant beating the clock. Get a headline ten seconds before the next trader, act on it, and you had an edge.

That edge is mostly gone now. High-frequency systems parse official wires in milliseconds, so by the time a headline reaches a normal screen, the price has usually already moved.

So where's the edge now? Increasingly, it's not in the news at all — it's in the exhaust trail of ordinary economic activity, showing up months before it ever reaches an earnings report or a government release. Researchers call this alternative data, and one strand of it has real, peer-reviewed backing: satellite images of shipping containers.

A 2023 study in Humanities and Social Sciences Communications (a Nature-portfolio journal) found that satellite-based container counts could predict stock returns in 27 of 33 countries studied, with a backtested annualized return of 16.38%. That's a real finding, not a marketing claim. Below is what the research actually shows, what's still just internet folklore, and how the different types of alternative data hold up.

What "alternative data" actually means

Alternative data is any non-traditional information — outside company financials, analyst notes, and government statistics — used to gauge economic activity or sentiment in something close to real time. Quarterly earnings look backward by definition. GDP and CPI numbers are worse: they're published with a lag and then revised, sometimes more than once, which limits how useful they are for predicting near-term returns. Alternative data, by contrast, can update daily or even hourly.

1. Satellite imagery — the best-documented source by far

Here's the study in question. A team of researchers (Yu, Hao, Wu, Zhao, and Wang) built a dataset of 83,672 satellite images covering the world's 48 largest container ports. They trained a U-Net deep learning model to count containers in each port from the imagery, using container volume as a stand-in for global shipping demand. The logic: containers piling up means demand is falling, and falling demand tends to show up in stock returns a bit later.

The results held up. The container-count signal significantly predicted daily stock index returns in 27 of 33 countries over 2019–2021, and a strategy built on that signal returned an annualized 16.38% in the backtest, with a Sharpe ratio of 1.19 — nearly double what a simple buy-and-hold approach produced over the same window. The effect was strongest during COVID, when port congestion became an unusually clean read on supply chains that weren't functioning normally.

This kind of imagery has other, less academically rigorous uses too. Hedge funds have been counting cars in Walmart, Target, and Costco parking lots since roughly the early 2010s to estimate foot traffic ahead of earnings — there's academic work (Zhu, 2019) showing satellite car counts can anticipate earnings surprises before they're announced. Firms like Orbital Insight and SpaceKnow track oil tank fill levels and industrial activity to get ahead of commodity moves, and multispectral imagery gets used to gauge crop health before official agriculture reports come out.

The catch: most commercial satellite data is priced for institutions, not individuals. And even with public imagery — the study used free Sentinel-2 data — you need real image-processing skill to turn pixels into a usable trading signal before someone else does it faster. That's not an accident. It's exactly why the edge hasn't been arbitraged away. It isn't a secret; it's a cost and skill barrier, which is a different thing.

2. Shipping and logistics data beyond the container counts

The same study found that its satellite-based signal led traditional shipping indicators — the Baltic Dry Index and container throughput indices — by about two months. That's notable because those traditional indicators are already considered leading indicators of industrial production. Getting two months ahead of a leading indicator is a meaningful head start.

Vessel-tracking systems like AIS are also used commercially to watch port congestion and ship movements. Some logistics intelligence firms look at vessel draft — how low a ship sits in the water, which indicates cargo weight — as an informal proxy for load. That approach is used in practice, but it's nowhere near as rigorously tested in peer-reviewed literature as the satellite container-count work.

3. Social media sentiment

Old-school sentiment analysis — tallying positive versus negative words in posts — has gotten less reliable as bots and coordinated posting campaigns have grown. The more recent academic work looks at graph-based sentiment analysis instead: not just what's being said, but who's saying it and how it spreads through a network. The network structure itself becomes part of the signal.

This is a legitimate and still-developing research area. But a lot of what circulates online about it — tracking specific emoji usage to catch pump-and-dump schemes, for example — is closer to pattern-spotting than published finding. Worth watching, not worth trading on as if it were settled science.

4. Transaction and web data

Providers like Second Measure sell anonymized credit card transaction data, and some funds track search volume trends, to estimate a company's revenue before its quarterly numbers come out — sometimes by weeks. This is standard practice among institutional investors. It's also licensed commercially and rarely available to individual traders at any scale that matters.

5. Corporate activity signals — treat with real skepticism

Private jet tracking and "pizza delivery spikes near corporate headquarters" as M&A tells make for good stories — the 2019 Occidental Petroleum jet-tracking anecdote gets cited constantly. But a single well-timed correlation isn't a tested, repeatable signal. This category is closer to entertainment than strategy, and nothing here has the kind of peer-reviewed backing the container-shipping research has.

Why would this edge exist at all, if markets are efficient?

Fair question. If markets price in all available information, shouldn't satellite data already be reflected in prices? The standard academic answer comes from the Grossman-Stiglitz framework on costly information: markets are only as efficient as the cost of getting the relevant information allows. Satellite processing, licensing, and skilled analysis all cost money, so the information gets priced in gradually, as more investors acquire and act on it, rather than all at once.

The study backs this up directly. It found the container-shipping signal took up to five trading days to fully work its way into prices. That lag is the cost-of-information dynamic showing up in the data.

What to actually take from this

The rigorously tested version of this story is narrower than the version that goes viral — but it's real. Satellite-based economic monitoring has peer-reviewed, statistically significant backing. Sentiment and transaction data have legitimate institutional uses, even where individual claims about them get overstated. And a few popular ideas — jet tracking, pizza deliveries, emoji velocity — are internet folklore dressed up as strategy.