Types of alternative data used by hedge funds, from card data to web data

The official descriptions cited below name four types of alternative data used by hedge funds: credit card and other transaction data, satellite imagery and location data, web and app usage, and sentiment and text. This post adds a fifth offered to funds, company-level web data, and maps each by what it observes, at what grain, how it joins to a security and which compliance questions it raises, so that you can place any dataset a vendor offers and know what to ask next. Nothing here is investment advice, and the post makes no claim that any dataset predicts a return. The glossary entry on alternative data holds the definition and the general checks.

Which types of alternative data are used by hedge funds, from credit card data to satellite imagery?

Four types of alternative data used by hedge funds appear in the official descriptions cited here: credit card and other transaction data, satellite imagery and phone location data, web traffic and app usage, and sentiment from social media and news text. This post adds a fifth, offered to funds: company-level web data such as job postings and website technologies.

The grouping builds on a discussion paper published by the European Central Bank in September 2026, which carries its authors' views and not the bank's. It sorts most alternative data into three categories: by-products of consumers' decisions, such as traffic on a firm's website, point-of-sale data and credit card data, then communications on social media, then sensor data such as satellite imagery or geolocation data. The table splits the first into purchases and usage, and its row for what a company publishes about itself is this post's own addition.

The fourth column is where most of the engineering lies. A dataset keyed on merchants, places or app names is resolved to companies and then to securities, as the guide to mapping alternative data to a security master sets out.

Alternative data in hedge fund investment research: satellite, credit card and web traffic data

Examples of alternative data in hedge fund investment research, in official descriptions, include satellite imagery of retailers' parking lots and crop fields, aggregate credit card transactions, web traffic and app usage estimates, social media sentiment and phone location data. Company-level web data, the fifth type below, is this post's addition and is offered to funds.

Credit card and other transaction data

Official bodies have used card data themselves, and their caveats are the questions to put to a vendor. The US Bureau of Economic Analysis built weekly and monthly estimates of spending by industry group from credit, debit and gift card transactions collected by a card intermediary, and no longer produces them: the last update was on 7 May 2024. Its page lists the limits: the data set is not necessarily representative of each industry, it covers primarily brick-and-mortar merchants, and merchants that leave the sample are not counted. Economists at the Federal Reserve Board built daily estimates of retail spending by geography from a payments company's anonymised transactions, in a staff working paper of August 2019 that gives its authors' conclusions and not the Board's.

Satellite imagery and location data

The Operational Land Imager on Landsat 9, a US Geological Survey and NASA satellite, has a maximum ground sampling distance of 30 metres, or 15 in its panchromatic band. The satellite repeats its path every 16 days. A 30-metre pixel covers 900 square metres, so ask a vendor of site counts for the resolution and revisit of its source imagery.

On 9 January 2024 the Federal Trade Commission announced its first settlement with a data broker over the collection and sale of sensitive location information. The order, proposed that day and made final on 12 April 2024, prohibits X-Mode Social and its successor Outlogic from sharing or selling sensitive location data, and requires a supplier assessment programme to ensure that their suppliers of location data obtain informed consent from consumers.

Web traffic and app usage

The figures here are estimates, so ask what went into them. On 14 September 2021 the Securities and Exchange Commission announced settled securities fraud charges against App Annie Inc., a seller of estimates of how often a company's mobile app is downloaded and used and how much revenue it generates. The release says the Commission's order found that the firm had assured companies their data would be aggregated and anonymised before a statistical model used it, and that from late 2014 through mid-2018 it used non-aggregated and non-anonymised data to alter its estimates. Without admitting or denying the findings, App Annie consented to a cease-and-desist order under which it was ordered to pay a penalty of $10 million, in what the release calls the SEC's first enforcement action charging an alternative data provider with securities fraud.

Sentiment and text

A report by FINRA staff dated December 2025, which gives its authors' views and not official views of FINRA, describes sentiment analysis tools that may use natural language processing and machine learning to parse social media posts, comments and discussions. Market participants from broker-dealers to hedge funds to asset managers use them, it says, to supplement traditional research. Some told the staff that the tools may struggle with sarcasm, idioms and misleading information, particularly across languages, and the report adds that their output depends on the quality and relevance of the input data. A first test follows: grade a sample of scored posts by hand, in each language you need.

Company web data: postings, technologies and announcements

The fifth type observes what a company does in public: the roles it opens, the technologies its website runs, the announcements and filings it makes. The company generates it, so it exists for a firm that sells only to other businesses as it does for a retailer. A posting records an intention to hire, and a detection shows presence on the website. The ECB paper above uses the type itself: its authors collected job postings from the career pages of 35 financial institutions to measure AI-related hiring. Twelve data protection authorities, the UK Information Commissioner's Office among them, stated on 24 August 2023 that publicly accessible personal information is still subject to data protection and privacy laws in most jurisdictions.

Fokals delivers this type, collected from first-party company sources and public records and processed in-house. Job Postings holds every role a company publishes. Technology Stack and Technology Changes hold the technologies each company runs, with first-seen and last-seen dates, and a dated event for every adoption and removal. Company News holds announcements and regulatory disclosures classified into 13 event types. It is company-level data throughout, every record carries the time it was observed, and a listed company carries its ticker, MIC, ISIN, LEI and share-class FIGI. The sourcing statement sets out what is collected and how.

Is there an SEC definition of alternative data for hedge funds and equity research?

A Risk Alert of 26 April 2022 from the SEC's Division of Examinations describes alternative data as the many kinds of information, beyond traditional financial statements, company filings and press releases, that are increasingly used in financial analysis. That is a staff description. We found no Commission rule defining the term in Title 17 of the Code of Federal Regulations.

Its examples are information drawn from satellite and drone imagery of crop fields and retailers' parking lots, analyses of aggregate credit card transactions, social media and internet search data, geolocation data from consumers' mobile phones, and email data from apps and tools that consumers may use. It adds that alternative data does not necessarily contain material non-public information. The alert represents the views of the staff, is not a rule, regulation or statement of the Commission and has no legal force or effect.

Section 204A of the Investment Advisers Act requires all investment advisers, registered and unregistered, to establish, maintain and enforce written policies and procedures reasonably designed to prevent the misuse of material non-public information. Staff observed advisers that used alternative data and did not appear to meet that requirement. Among the alert's examples:

• diligence of data providers that was ad hoc, inconsistent or not written down
• no policy for assessing the terms, conditions or legal obligations attached to how the data was collected or provided
• no system for deciding when to repeat the review of a provider, as time passes or its collection practices change

The Commission's own release of September 2021, cited above, says of estimates of mobile app performance that trading firms commonly call this type of information alternative data because it is not contained in companies' financial statements or other traditional data sources.

In the UK, the Financial Conduct Authority's feedback statement of January 2022 gives a high-level definition: non-traditional data generated outside issuing firms, typically gathered by intermediaries such as data vendors, and often a by-product of another business process. Its examples are satellite imagery data, account-level data, location data, credit card transaction data and social media. By the SEC staff's description company filings and press releases are traditional, and by the FCA's definition alternative data is generated outside issuing firms, so the table's fifth row, which includes a company's announcements and filings, is this post's grouping and not part of either body's description.

Which alternative data types have the highest adoption among asset managers?

We found no official figure for which alternative data types have the highest adoption among asset managers. We searched the publications of regulators and central banks in the United States, the United Kingdom and the European Union and found none that gives adoption by type, so this post gives no ranking.

The FCA's feedback statement says data and advanced analytics are increasingly used by a wide range of market participants, including hedge funds, investment managers, banks and exchanges, and that most respondents named cost as the biggest barrier to using alternative data. The FCA added that it had commissioned research on the nature and scale of alternative data usage, and on the FCA's site we found no figures by type published from it.

The nearest official count we found is about artificial intelligence. A risk analysis published by the European Securities and Markets Authority on 20 February 2026 reports a survey by the Dutch Authority for the Financial Markets in the first half of 2025, answered by 323 entities: 173 investment firms, 95 fund managers, 36 proprietary traders and 19 other financial firms. Analysis of unstructured, alternative or big data was the second most cited use of AI, with 81 mentions. That counts uses of AI at Dutch firms and says nothing on which type of data they buy.

Which alternative data should a hedge fund buy first: credit card data, web scraping or sentiment?

No type of alternative data comes first for every hedge fund. Credit card data, web-scraped data and sentiment data inform different questions, so the first purchase follows from the questions a fund asks, the companies it holds, the work of mapping each dataset to securities and the compliance review each source needs.

This post ranks no type above another and names no vendor to buy from. Five steps put the types in order for one team.

1. Write the questions down, and beside each the companies it applies to.
2. Mark which type observes each question.
3. Estimate the work of bringing each type in, in analyst-weeks: the join to your security master, the history check and the tests.
4. Put each source through a written compliance review before a trial, with a due diligence questionnaire, and set the date of the next review.
5. Test a sample on the companies you hold, using each value only from the date it was observed, which a point-in-time record allows.

Worked example: alternative data priorities for a small hedge fund, from transaction data to web data and sentiment data

A small hedge fund can set its alternative data priorities by counting, for each type, the questions it informs across the companies held, and dividing by the weeks of work to bring it in. In this illustrative example company web data comes first for one book and transaction data for another. The order belongs to the assumptions.

Assumptions:

• The team has 12 analyst-weeks a year for bringing in new data.
• The book holds 40 listed companies: 10 sell to consumers and 30 to businesses. Of the 30, 12 are software vendors and one of the other 18 is Acme Robotics, an illustrative maker of warehouse robots.
• Transaction data informs two questions on the 10 consumer companies: whether sales this quarter run ahead of the last, and whether spending is moving between competitors.
• Company web data informs two questions: whether a company is adding roles or markets before it reports them, on all 40, and whether customers are adopting or removing a vendor's product, on the 12 software vendors.
• Sentiment data informs one question, whether public discussion turned after an event, on the 10 consumer companies and the 12 software vendors.
• Bringing in a type takes 8 analyst-weeks for transaction data, 4 for company web data and 5 for sentiment data.
• A type's score is its company-questions divided by its weeks.

By score the order is company web data, sentiment data, transaction data. The first two take 4 + 5 = 9 weeks, and the 3 weeks left are fewer than the 8 that transaction data needs, so it waits a year. For Acme Robotics only the question about roles and markets applies.

Now make all 40 companies consumer companies. The two transaction questions apply to all 40 and the team adds a third, whether growth comes from more buyers or larger baskets. The software question applies to none, and the sentiment question to all 40.

The order becomes transaction data, company web data, sentiment data. The first two take 8 + 4 = 12 weeks, the whole year, so sentiment data waits. Neither order is advice: change a question, a count or an effort and the order moves.

Frequently asked questions

What this page says about the products it names was checked against their public documentation on 7 October 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.