AI glossary · Basic · Data engineering · Finance & insurance
Alternative data
Alternative data is information obtained outside traditional financial data such as financial statements, stock prices and credit records. It means using sources such as card payment flows, satellite images, web traffic and phone bill payment history for investment decisions or credit scoring.
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Industry AI
- Fields
- Data engineering
- Industries
- Finance & insurance
- Added
- 2026-10-10
How it works
Traditional data arrives late and in a fixed format, like financial statements released every quarter. Alternative data such as payment records, app usage, online reviews, job postings and satellite imagery accumulates more often and in more varied forms. This data is cleaned and analyzed with machine learning to turn it into signals that give an early estimate of a company's revenue or an individual's ability to repay. Unstructured data such as satellite imagery or review text first has to be converted into numbers or categories using AI.
Traditional data vs. alternative data
Traditional dataFinancial statements, stock prices, credit records
- Comes in a fixed format
- Arrives late, e.g. quarterly
- Long track record of proven standards
Alternative dataPayments, satellite imagery, web traffic, etc.
- Accumulates more often, in varied forms
- Needs cleaning and AI analysis first
- Needs privacy and fairness checks
공통 Both are used to estimate company performance or an individual's ability to repay
Why it matters
In investing, it is used to read a company's situation before official earnings announcements and make decisions faster than others. In credit scoring, it can widen access to finance by assessing the repayment ability of people who find it hard to get a score because they have almost no loan or card history, such as young people just starting work or self-employed business owners, using phone bill and utility payment history, revenue flows and similar data. As AI became able to process large amounts of unstructured data, the range of uses has expanded quickly.
Things to keep in mind
Using personal data requires the person's consent and compliance with privacy regulations, and if de-identification is not done properly, there is a risk of re-identification. Information such as where someone lives or what device they use can indirectly reveal sensitive characteristics such as gender or income and lead to discriminatory lending decisions, so financial regulators require fairness and explainability. If the data represents only some groups, or the collection method changes partway through, the signal becomes distorted. In investing, firms also need to check that the way the data was obtained does not raise legal problems such as the use of material non-public information.
Example
One asset management firm counts cars in satellite images of large supermarket parking lots and uses this as a reference for estimating the retailer's revenue trend before its earnings announcement. One internet-only bank includes phone bill payment history and payment data the customer has consented to share when reviewing loans for customers with a short financial history, giving loan opportunities to people who would have been rejected based on a traditional credit score alone. In both cases, the data source, consent and the model's fairness are checked separately.