Facebook Data Mining
Facebook is one of the largest single sources of human behavioural data ever assembled. People use it to stay in touch with friends and family, share photos and videos, and follow news; businesses run pages to promote products and reach audiences; bloggers and influencers use it to build followings and drive engagement through likes, comments, and shares. For many users, checking Facebook is simply a daily habit.
That combination — billions of people, using the platform for many different purposes, every day — is what makes Facebook interesting to data mining. Every interaction leaves a record, and those records add up to a dataset that can be analyzed for patterns in how people behave, connect, and communicate.
Facebook at a Glance
| Launched | February 2004, at Harvard University |
| Founders | Mark Zuckerberg, with Eduardo Saverin, Andrew McCollum, Dustin Moskovitz, and Chris Hughes |
| Headquarters | Menlo Park, California, United States |
| Parent company | Meta Platforms (the company renamed itself from Facebook, Inc. in October 2021) |
| Minimum age | 13 in most countries (higher in some jurisdictions) |
Scale figures move constantly, so treat any specific number as a snapshot rather than a fixed fact. As of Meta's reporting in early 2026, Facebook had roughly 3.07 billion monthly active users and 2.11 billion daily active users — meaning close to seven in ten monthly users open the platform on any given day.
One figure worth correcting, because it's widely misquoted: Facebook's global audience skews male, at roughly 56% male to 43% female according to Meta's own advertising audience data. The often-repeated "more women than men" statistic comes from the United States specifically, where the split runs about 54% female to 46% male. Global and national figures are not interchangeable, and this is a good reminder to check which population a statistic actually describes before repeating it.
What Data Facebook Holds
Users voluntarily supply a great deal of structured personal information: date of birth, education history, employment details, hobbies, interests, relationship status, and location. On top of that sits everything generated by using the platform — photos, posts, comments, messages, reactions, group memberships, page follows, and the friend connections that link every account to others.
That last category matters most for data mining. Profile fields describe individuals; the connections between accounts describe a social network, and network structure supports a whole class of analysis that isolated records cannot.
What Data Mining Actually Does With It
"Analyzing Facebook data" covers several distinct techniques, most of which appear elsewhere in this series:
- Community detection — a form of clustering applied to the friendship graph, used to find tightly connected groups of people who may not be formally labeled as a group anywhere on the platform.
- Sentiment analysis — text mining applied to posts and comments to gauge whether opinion about a topic, product, or event is positive or negative.
- Topic detection and tracking — identifying what subjects are being discussed and how that shifts over time.
- Link prediction — estimating which two users are likely to connect next, the technique behind "people you may know" suggestions.
- Audience segmentation and ad targeting — clustering users by interests and behaviour so advertisers can reach specific groups. This is the commercial engine of the platform, not an academic sideline.
Academic researchers have used Facebook data to study the social networks of college students and how people interact online, and comparable studies on other platforms examine how users express feelings and opinions in public. The techniques are the same ones covered earlier in this series; only the data source is unusual.
Why Access Is Restricted Today
A lesson written a decade ago would have described Facebook as an open, easily-mined data source. That is no longer accurate, and understanding why is part of understanding the subject.
In 2018, the Cambridge Analytica episode revealed that a personality-quiz app had harvested data not just from its users but from their friends as well, and that this data had been passed to a political consulting firm. The fallout reshaped the entire field: Facebook sharply restricted what its Graph API exposes about users and their connections, and other platforms tightened access in parallel.
Access has continued to narrow since. CrowdTangle, the tool many journalists and academics relied on to study public Facebook and Instagram content, was shut down by Meta in August 2024 and replaced by the Meta Content Library, which researchers have widely criticised as harder to access and less transparent than what it replaced.
The practical consequence: large-scale Facebook data mining today is mostly done by Meta on its own data, or by approved researchers working within controlled access programs. Scraping profile data without permission violates the platform's terms of service, and in many jurisdictions also runs into data protection law.
Ethics, Approval, and the Law
Because this data describes real, identifiable people, collecting and analyzing it raises ethical and privacy obligations that go beyond what is technically possible.
Institutional review. Universities operate Institutional Review Boards (IRBs) that review research involving human participants. Under U.S. Department of Health and Human Services regulations, federally-supported research involving human subjects must be reviewed and approved by an IRB before it begins — rules designed to ensure participation is voluntary, risks are minimised, and individuals' rights and privacy are protected.
It's worth being precise about scope, though, because this is easy to get wrong: IRB review governs institutional research. A marketing analyst at a company studying their own page's audience is not covered by an IRB at all. What governs them instead is the platform's terms of service and applicable data protection law — the GDPR in the European Union, and comparable regimes elsewhere — which apply regardless of whether the work is called research.
Public does not mean unrestricted. A post being visible to anyone does not make it free to collect, store, and republish. Users share information in a particular context and generally do not expect it to be assembled into a research dataset, and re-identification is a genuine risk: even data stripped of names can often be traced back to individuals when combined with other information.
The short version: on Facebook, the ethical and legal constraints are usually tighter than the technical ones. Anyone mining this kind of data needs to establish what they're permitted to collect before asking what they're able to collect.
Related Concepts
The techniques listed above — clustering, text mining, and classification — are covered in this series' lesson on data mining techniques, and the graph-based methods used on friend networks are covered in the lesson on social media data mining methods.