My Social Graph is Broken So I Have No Idea Who My Friends Are

In 2010, Facebook launched Places. It was their answer to Foursquare.

It failed almost immediately—not because the feature was bad.

The graph was wrong.

When people built their Foursquare networks, they did it knowing exactly what the app was for. A Foursquare friend request meant something very specific: You’d be ok running into this person IRL. Facebook's graph was built for something else entirely—reconnecting with classmates, staying in touch with family, documenting life milestones. By the time Facebook tried to bolt location sharing onto that, nobody wanted to tell their grandmother where they were having drinks on a Tuesday.

The graph had no context. It failed.

I've been thinking about this a lot lately, partly because I've been poking around Goodword’s relationship management platform and a handful of other "network management" apps. Also, I was talking to Vic Singh, who is building Originalis—an operating system for venture capital.

He said something that crystallized the whole problem.

Vic was explaining how Originalis uses network intelligence to help VCs figure out who they should pull in for diligence on a deal, who they should be warming relationships with before a fundraise, and who the right follow-on investors are when a portfolio company goes out for a Series B. What he kept coming back to was: you can't do any of that if the underlying graph doesn't know what the relationships actually mean. You might have a connection to someone on LinkedIn, but does Originalis know that you met them once at a conference versus that they were your co-founder for three years? Those are not the same relationship. The strength, the recency, the context—all of it matters.

Without that, you're not working with a network. You're working with a contact list.

That's the problem every consumer network management app has failed to solve, and most of them haven't even tried. There are lots of other apps that have asked for my network as well—to help me figure out who to share the app with.

Here's what typically happens: you download one of these apps and you start alone. You need connections to make it worthwhile, but you don’t know who you know that’s already on. In the early days, you might be the first one through the door.

It asks you to connect your LinkedIn, Facebook, etc., or upload your contacts. You're then handed a decontextualized blob of everyone you've ever emailed or connected with—your college roommate, the person you met at a conference in 2017, your kid's pediatrician, a founder you passed on, your high school friend who sells insurance now. They're all in there.

There's no single player mode. There's no foundation. And there's no way to act intelligently on a graph without understanding what each relationship actually is.

What makes this harder is that "network management" isn't even one problem. For me, right now, it's at least five distinct ones: making more time for individual friends when I'm a time-poor parent of a toddler; building a lightweight professional CRM for non-sales relationship work like VC coaching, finding activity partners (local cyclists, softball players who can rake), trying to find more couple friends in Brooklyn with similarly aged kids who also have a Wednesday date night. These all involve discovery, scheduling, and accountability at different times.

Those are not the same problem. An app that tries to solve all at once will likely solve none of them well.

What would it actually take to do this right? If you’re really going to be the next LinkedIn or the world’s best personal CRM, what’s the first hurdle?

I’d start by helping people understand who they actually know before you try to do anything with that information. Run a structured interview. Superhuman scaled plenty while still requiring a 1:1 chat to walk you through their app, so we’ve seen it work before when someone considers the problem important enough and the system promising.

Walk a user through their circles. Where did you grow up, and do you still talk to anyone from then? Walk me through your resume—what communities did each chapter create? Where does your kid go to school, and how do those parents communicate? What WhatsApp groups are you in? Layer in cross-platform data—calendar, email, some entity resolution to figure out that OldMetsFan1986 on Instagram is actually my friend Bob.

Then, critically, ask what success looks like in each circle. Not "do you want deal flow?"—that's a LinkedIn question. Ask: what does being a good friend look like to you? What do you regret not doing better?

That messy, friction-heavy onboarding is the product. The features everyone's building on top of it are only valuable if that foundation is there. Vic had to fight for this at Originalis. In his first board meeting, one of his investors told him nobody had ever cracked the network intelligence problem and he should skip it.

He didn't, because he knew the whole thing falls apart without it.

Most consumer apps skip it because it doesn't scale the way investors want things to scale. So we keep getting apps that ask for your whole contact list, build features on top of a broken graph, and then wonder why nobody sticks around or finds value.

Facebook Places was sixteen years ago and we still haven't learned this lesson.