The Advice Everyone Gives
When a job search stalls, the advice arrives in one familiar form: call your closest people. Phone the former manager who knows your work, the friend who would vouch for you, the colleague who has seen you under pressure. The logic feels airtight: the people who care most about you should try hardest for you.
That advice is wrong in a specific, measurable way. Acquaintances, not confidants.
A team from LinkedIn, Harvard, MIT, and Stanford spent five years testing the oldest idea in network sociology with an experiment rather than a survey, and the experiment did not flatter the inner circle. The people most likely to lead you to a new employer were not your closest friends, but acquaintances: people you share a handful of friends with, who hear about openings your own circle never encounters.
An Experiment, Not a Survey
The distinction matters, because almost everything written about networking before this study rested on people remembering how they found jobs, and memory in that setting is a rigged witness. People remember the friend who made the call and forget the acquaintance whose casual mention put the opening on their radar.
This study changed the mechanism itself. LinkedIn's People You May Know feature recommends new connections, and the researchers used randomized variations of that engine to nudge some users toward more weak ties and others toward more strong ties. Because the nudge was random, users who ended up with different networks did not differ systematically in ambition or talent, which is what lets the authors speak about cause rather than correlation. They then tracked job transmissions under the strict timing rules in the Study Card, designed to weed out coincidences at the same large company.
The scale is hard to dismiss: more than 20 million people, roughly 2 billion new ties, and 600,000 new jobs from 2015 to 2019, answering an objection that has hung over this theory since Granovetter first stated it in 1973.
The Peak at Ten Mutual Friends
The headline finding needs its twist, because the study does not say that weaker is always better. On the measure that counts shared connections, the relationship between tie strength and job transmission traced an inverted U: transmission rose as ties weakened from the strongest friendships, peaked at moderately weak ties, and then fell again. The peak sat at about 10 mutual connections, describing an acquaintance in a partly overlapping world.
Why does that shape appear? Your closest friends tend to know the same people you know and hear about the same openings, so a lead from them is often a lead you already have. A total stranger may hold entirely new information but has little reason to trust you with it and no mutual friend to make the introduction feel safe. The moderately weak tie sits between those failures, with just enough shared connections to move information and just enough unshared connections to make it new.
The second measure complicates the picture. When weakness was measured by how rarely two people exchanged direct messages, rather than by shared connections, the weakest ties produced the most mobility, with no peak in the middle. Two definitions of a weak tie gave two different shapes, and industry split the result a third way, with weak ties winning most in more digital industries and strong ties winning in less digital ones.
The Arithmetic Nobody Did
Start with the raw scale, because the study's numbers support a calculation the press coverage never ran. Two billion new ties divided by 20 million people is 100 new ties per person on average, and 600,000 new jobs divided by 20 million people is 3 new jobs per 100 people during the experimental window. Most transmitted nothing, which is the base rate against which the tie effect should be judged: transmission through any single new connection is rare, which is why the composition of those connections matters more than the raw count.
Now apply the reported effect. If a moderately weak tie transmits a job at more than twice the rate of a strong tie, then swapping 10 strong-tie additions for 10 moderately weak additions more than doubles the expected transmissions from that batch of 10, even though the total number of connections has not changed. Indiscriminate collecting does not buy 100 chances at the peak rate, because the peak sits in the middle.
The Strongest Counterargument
The strongest case against this finding starts with its own outcome measure and does not let go, because job transmission means joining the same employer as a connection under timing rules, and that is not the same as proving the connection caused the job, since two people in the same industry, city, or alumni network can plausibly end up at the same employer through shared background alone, and the timing rules reduce that problem without eliminating it, and the randomization changed recommendations rather than friendships themselves, so the claim rests on the assumption that the nudge worked only through the ties it created. That assumption is the study's softest spot, because a recommendation nudge can also make users feel noticed, browse more job postings, and apply more widely, and any of those behaviors would lift job mobility without a single new tie doing any work.
A second challenge comes from what strong ties do best, which this outcome never measured. Complex, high-trust help, the kind where someone stakes their reputation on you or vouches for you in a closed meeting, is exactly where close relationships should win, and earlier work on complex contagions found that behaviors requiring social reinforcement spread better through clustered, strong ties than through long, weak bridges. A weak tie can deliver a job posting, while a strong tie can deliver a sponsor.
The third challenge is the most practical. LinkedIn users are not the workforce. They skew professional, digital, and already connected, the experiments ran from 2015 to 2019 in a pre-pandemic hiring market, and the authors include LinkedIn employees working with data that no outside team can freely re-analyze. A result this large and this hard to reproduce independently deserves to be treated as a strong lead rather than a settled law.
What We Didn't Prove
This evidence does not prove that collecting acquaintances will get you a specific job, because job transmission is a population-level probability and most new ties transmitted nothing at all. It does not prove the finding travels beyond LinkedIn, beyond professional and digital work, or beyond the 2015 to 2019 hiring market. It does not prove weak ties improve wages, satisfaction, or promotion speed, because none of those were the study's measure. It does not prove the peak at 10 mutual connections is a universal constant, since that number comes from one platform at one point in time. Finally, it does not prove strong ties are useless: in less digital industries they won.
The Bottom Line
Your closest friends are the people most willing to help you, and they are often the worst positioned to tell you something new. The middle of your network is where new information enters. The LinkedIn experiment gives that old tension a causal test, a curve, and a peak: moderately weak ties, around 10 mutual connections, carried job information further than either close friends or near-strangers, by more than double the strong-tie rate.
What You Can Do
First, audit the middle of your network rather than its edges. Look for people you have not spoken to in a year who share roughly a handful to a dozen mutual connections with you, especially former colleagues, classmates, and conference acquaintances in adjacent fields.
Second, make the re-engagement specific and low-pressure. Send one short message that names the shared connection or event, states what kind of work you are looking for, and asks one concrete question, such as who else in their world is hiring for that kind of role.
Third, build new moderately weak ties before you need them. Attend one event slightly outside your specialty each quarter, accept connection requests from people you met briefly, and follow up within a week while the shared context is still fresh. In less digital industries, keep investing in strong ties too, because this study found they still won there.