How Facebook Picks Your People You May Know
Each time you log into Facebook, you might notice familiar or not-so-familiar faces under the “People You May Know” section. These suggestions often spark curiosity: How does Facebook know these people? Why are some suggestions so accurate—and others seemingly random? While Facebook does not reveal every detail of its proprietary algorithms, enough information has been provided over the years, and researchers have closely studied its patterns to allow us to build a clear picture of how your potential connections are determined.
The Purpose Behind the Feature
The “People You May Know” (PYMK) feature is not just a convenience—it’s a core part of Facebook’s growth and engagement strategy. Encouraging users to connect with more people increases user interaction, data sharing, time spent on the platform, and ultimately, advertising revenue.
This is why understanding how Facebook chooses who to suggest can also provide insight into the platform’s sophisticated data mining and predictive analytics capabilities.
How Facebook Gathers Relationship Clues
Facebook employs a range of data signals to build a profile of your network and suggest new connections. Here’s a breakdown of the most significant sources:
- Mutual Friends: The more mutual friends you share, the more likely that person will appear in your suggestions. This is the most obvious and visible factor.
- Contacts Uploaded: If you have allowed Facebook to access your phone contacts or email address book, the platform uses this data to suggest connections.
- Groups and Pages: Facebook notes whether two users are part of the same groups, follow the same pages, or attend the same events.
- Shared Workplaces or Education: Listing the same school or employer—even if you didn’t attend at the same time—can be a determining factor.
- Profile Interactions: If you’ve viewed someone’s profile or liked similar types of posts and content, that behavioral data could trigger a connection suggestion.
The Role of Advanced Algorithms
Behind the scenes, Facebook uses machine learning models that constantly update their predictions based on new data inputs. These dynamic models analyze trillions of interactions across millions of users each day.
Graph theory lies at the heart of the process, where users are nodes and their interactions are edges connecting them. Facebook builds a dynamic and evolving social graph to analyze how tightly interconnected users are to others beyond their immediate friend list.
The more edges (shared traits, interactions, similarities) between two nodes, the higher the probability that a suggestion will be made. These suggestions are calculated probabilistically, meaning a suggested friend might not be someone you know today—but someone you’re likely to meet or interact with in the near future.
Location and Proximity Data
In many cases, location data plays a crucial—if controversial—role. Though Facebook has stated it doesn’t use GPS information from your phone explicitly to suggest friends, multiple reports and user experiences have led some to believe it does so indirectly.
For example, if you and another user attend the same event, are in the same café, or are within Bluetooth or Wi-Fi range, Facebook might interpret this proximity as a potential social link, even if you haven’t shared any mutual friends.
Facebook has responded that it uses “location data as one of many signals” and never relies on location alone to make friend suggestions. But given its access to detailed geographic information—including check-ins, geotagged photos, and IP addresses—it’s reasonable to conclude that location contributes in a nuanced and indirect way.
Third-Party App Data and Shadow Profiles
This category is perhaps the most opaque—and the most contentious—in Facebook’s ecosystem. Facebook collects data not just from direct user actions, but also indirectly through shadow profiles. These consist of inferred data about non-users or about data not intentionally shared.
If a friend uploads their contact list and that list contains your number—even if you never gave Facebook your contact list—you can still end up being suggested to mutual friends. This complex web of data-sharing has led privacy advocates to criticize Facebook’s practices as borderline invasive.
This phenomenon becomes even more apparent when users receive PYMK suggestions for people they have only emailed or messaged privately—suggesting a level of backend cross-referencing far beyond simple social overlap.
Your Own Viewing Behavior
Every action you take on Facebook—browsing a stranger’s profile, clicking on photos, or searching someone’s name—can influence your suggestions. Facebook’s algorithm interprets engagement as interest. That means if you view someone’s profile several times, and they also happen to be linked to your extended network, the system may push them into your PYMK because of this behavioral signal.
Equally, if someone checks your profile, they might begin seeing you as a suggestion. While Facebook doesn’t officially confirm who views whom, this mutual interaction trail can quietly work in both directions.
Integration with Instagram and Other Platforms
Since Facebook also owns Instagram and WhatsApp, cross-platform data sharing is part of the friend suggestion infrastructure. If you follow someone on Instagram, and they also link their account to Facebook, there’s a high probability they will show up in your PYMK feed—even if you have no interactions within the Facebook app itself.
This integration allows Facebook to unify user identities across platforms and enhance its predictive capability—even if users believe their social presences are compartmentalized.
Why You Might See Unexpected Suggestions
Many people report seeing bizarre or even unsettling friend suggestions—such as old acquaintances, strangers they walked past, or professional contacts they’ve never officially exchanged information with.
Here are a few explanations for this:
- Contacts uploaded by others: If several of your contacts have saved your email or phone number, even without your consent, Facebook compiles that data as a part of your extended identity network.
- Background data matching: The platform may infer a connection based on secondary relationships or shared data points that are too subtle for users to notice.
- Testing new models: Facebook continuously tests experimental algorithms, and some inaccurate suggestions may be part of such iterations.
How to Take Control Over Suggestions
While there’s no way to completely opt out of the PYMK feature, you can limit its accuracy and relevance by adjusting your settings:
- Disable contact uploading: On both the app and browser, you can remove previously uploaded contacts and block Facebook from future access.
- Limit profile visibility: Restrict who can see your friend list or personal information to reduce indirect linkages.
- Clear location history: In Facebook’s settings, you can disable location tracking and delete stored location data.
These steps won’t stop PYMK suggestions entirely, but they can reduce the number of cross-signal suggestions.
Conclusion
The “People You May Know” feature is far from arbitrary. It is the product of sophisticated algorithms, behavioral tracking, and undeniable data intelligence. By combining network structure, device data, and user behavior, Facebook strives to anticipate your social intentions before you ever act on them.
Whether you view it as a tool for reconnecting or a concerning use of personal data, understanding how PYMK works gives you greater insight into one of the most advanced social algorithms in operation today.