Alternative Data in Finance Explained
Satellite images, card swipes, and web traffic aren't traditionally financial data — but processed at scale, they can move ahead of a company's own quarterly numbers.
What counts as "alternative"
Alternative data is a catch-all term for any dataset that isn't a traditional financial source — company filings, analyst reports, or price and volume history. Common categories include satellite imagery (counting cars in retail parking lots, tracking oil storage tank levels), aggregated and anonymized credit card transaction data, web traffic and app usage statistics, shipping and logistics records, and social media activity. None of it was collected with investing in mind, which is exactly why it can carry information the market hasn't priced in yet.
Why AI made it usable
These datasets are enormous and mostly unstructured — millions of satellite images, billions of anonymized transactions, terabytes of web logs. A human analyst can't sift through raw satellite photos looking for parking lot occupancy trends across thousands of store locations. Machine learning models can process that volume automatically, converting raw images or transaction logs into a structured, trackable metric — an estimated quarter-over-quarter sales trend, for instance — that an analyst can actually use.
How the signal gets built
The typical path runs from raw data to a usable estimate: a satellite image gets processed by a computer vision model trained to count vehicles; that count gets aggregated across a retailer's stores and compared to prior periods; the resulting trend gets compared against analyst consensus forecasts for the same quarter. Similar pipelines exist for web traffic (estimating engagement from page views) and transaction data (estimating consumer spend by merchant category), each translating raw signal into an investment-relevant number.
What it's actually used for
- Earlier reads on company performance. Card-spend and web-traffic data can suggest how a quarter is trending before the company reports results.
- Supply chain and macro tracking. Shipping and satellite data can indicate demand or supply shifts in commodities and industrials ahead of official statistics.
- Cross-checking management commentary. Independent data can be compared against what a company says on an earnings call to test whether the narrative matches observable activity.
The limits worth knowing
Alternative data is noisy and indirect — a parking lot count is a proxy for sales, not sales itself, and proxies can be wrong for reasons unrelated to the business (weather, seasonal quirks, a store remodel). It's also not free or evenly distributed: large funds were early adopters because of the cost of acquiring and processing it, though vendors increasingly sell pre-processed signals to a wider range of institutional investors, narrowing that gap over time.
Alternative data is one of the inputs prediction markets themselves can draw on — see Prediction Markets and Alternative Data →
Quick answers
What is alternative data in investing?
Any dataset outside traditional sources like filings, price history, and analyst reports. Examples include satellite imagery, aggregated card transaction data, web traffic, app usage, shipping records, and social media activity.
Why does alternative data require AI to be useful?
Most alternative datasets are unstructured or enormous in volume, and only became usable for investment research once machine learning could process that scale and extract structured signals automatically.
Is alternative data only used by large hedge funds?
Large funds were early adopters given the cost of acquiring and processing these datasets, but processed signals are increasingly sold by vendors to a broader range of institutional investors, narrowing the access gap over time.