Alternative data is data from outside a company's own reports that investors use to estimate how the company is doing before it announces its results. Examples include anonymised card transactions, app downloads, website visits, job postings, satellite photos of car parks and prices copied from shopping websites. It is called alternative because it is not the traditional data that investors use: a company's filings, its share price and analysts' estimates. This page uses the United States as its example, because most buyers are US investment funds.
Why do funds pay for it? A listed company in the US reports its sales once every three months. An investor who estimates those sales correctly before the announcement can buy or sell shares ahead of others. So funds buy datasets that rise and fall with a company's sales.
The most valuable kind is card transaction data. A data company licenses anonymised records from a panel of millions of credit and debit cards, with names and card numbers removed. Each record shows an amount, a date and the shop where it was spent. The company cleans the shop names, because one coffee chain may appear in hundreds of forms. It links each shop to the listed company that owns it. Then it corrects for the panel's bias, because the panel's cardholders are not exactly like all Americans.
Imagine an analyst at an investment fund in New York who follows a US restaurant chain. Halfway through the quarter, the card panel shows that spending at the chain is 6% higher than a year earlier. Other analysts expect 3% growth, on average. App and website data also show more online orders. The fund buys more of the chain's shares before the results. When the chain reports growth of 5.5%, its share price rises. The data did not make the decision for the fund. It gave the fund a better estimate, earlier than others.
Other kinds of alternative data come from web scraping, global trade data and labour market analytics, which counts job postings. The datasets reach analysts as files loaded into their own systems, or through financial research platforms.
The rules are strict. In the US, it is illegal to trade on important information that is not public and was obtained in breach of a duty. So a fund checks that every dataset was collected legally, and that no record can be traced back to an individual person. Laws such as those explained in data protection law also apply.
How is it different from a financial data terminal? A terminal holds the traditional data: prices, filings and estimates. Alternative data is extra evidence, gathered from outside the company.