“70% of jobs are never advertised” is one of the most repeated statistics around, and nobody can ever point to its source. You won’t find it here either.
What does exist is something far sturdier: a causal experiment on twenty million people, published in Science in 2022, that measured which contacts genuinely raise the odds of changing jobs. The result is counterintuitive and has immediate practical consequences.
The original hypothesis, from 1973
Mark Granovetter published in the American Journal of Sociology a paper that became one of sociology’s most cited: The Strength of Weak Ties.
The idea is simple. The people closest to you (family, close friends, daily colleagues) largely move in the same circles you do. So they know roughly what you know. Casual acquaintances, by contrast, move in different circuits: they are the only channel through which information you couldn’t otherwise get reaches you.
In the fieldwork accompanying the theory, Granovetter interviewed professionals and managers who had changed jobs. Among those who found the position through a personal contact, only a minority had got the tip from someone they saw often: the majority came from people they saw occasionally or rarely.
The limitation of that work, acknowledged for decades, is that it was correlational. It could be that people with many weak ties simply differ in other ways.
The experiment that actually tested it
In 2022 Rajkumar, Saint-Jacques, Bojinov, Brynjolfsson and Aral published in Science a paper with an explicit title: A causal test of the strength of weak ties.
The design exploits years of experiments on LinkedIn’s “People You May Know” algorithm, which varied in a controlled way the type of contacts suggested to around 20 million users over five years, producing billions of new connections and hundreds of thousands of traceable job moves.
The result:
The ties that most increase the probability of changing jobs are neither the strongest nor the weakest: they are the moderately weak ones, with few mutual connections. The strongest ties turn out to be the least useful for job mobility.
It is causal confirmation of the 1973 intuition, with a useful refinement: the relationship isn’t linear. A complete stranger is no use; a lifelong friend isn’t either. What works is the person in between, someone you know well enough to message but who lives in a different world from yours.
A methodological note worth stating: the study measures job mobility on a professional platform, among users of that platform. It is not the whole workforce, and the result varies by sector: in more digital fields the weak-tie effect is more pronounced.
Why referrals genuinely matter
A second strand of research explains why companies favour this channel. Burks, Cowgill, Hoffman and Housman (Quarterly Journal of Economics, 2015) analysed large-scale hiring data comparing referred and non-referred candidates.
The main findings:
- referred candidates have a markedly higher probability of being hired at the same stage of the process;
- once hired, they show similar productivity to non-referred hires;
- but they stay longer, which makes them more profitable overall for the employer.
The practical implication is less romantic than usually told: a referral does not make you look talented. It makes you look less risky. Your CV is still necessary (we have a dedicated guide on writing one), but it stops being the only filter.
Why it makes you feel dirty (there’s a study on that too)
If the idea of “networking” makes you physically uncomfortable, that isn’t a character flaw: it is a documented effect.
Casciaro, Gino and Kouchaki (Administrative Science Quarterly, 2014) showed across a series of experiments that instrumental networking, the kind explicitly aimed at professional gain, produces a sense of moral impurity in participants, to the point of increasing the desire for physical cleansing. Spontaneous networking, born of curiosity or affinity, does not produce the same effect.
The most interesting detail is who suffers least: people in positions of power. Those who already have a lot to offer don’t experience the exchange as instrumental, because it isn’t.
That points to the only workable solution for people in ordinary positions: have something to give before you have something to ask. This isn’t moral advice, it is the removal of the condition that produces the discomfort.
Structural holes, or where the value sits
Ronald Burt (American Journal of Sociology, 2004) added the missing piece: what matters isn’t only how many contacts you have, it is where you sit in the network.
A “structural hole” is the empty space between two groups that don’t talk to each other. Whoever straddles that gap, say someone who knows people in engineering and in sales, has access to information neither group holds alone. In Burt’s data, that position is associated with better evaluations, ideas judged more valuable, and higher pay.
The operational translation is precise: five contacts in five different worlds are worth more than twenty-five in the same one. If 90% of your network does your job in your city, you don’t have a network: you have an echo.
Five moves for a network that works
1. Map your weak ties, not your friends. Take half an hour and write down the names of people you know but haven’t spoken to in over a year: former colleagues, former classmates, people met on a project. That is precisely the band identified by the Science study.
2. Reactivate, don’t introduce yourself. The message that works has a specific reference and no request: “I saw you moved to X, I was working on that too. How did it go?”. Not “hi, how are you, I’m writing because I’m looking for work”.
3. Ask for information, not favours. “How does hiring work at your place?” is a question almost everyone answers. “Can you refer me?” puts the person in an awkward spot if they don’t know you well enough.
4. Ask for more than you think. Vanessa Bohns’s research on the underestimation-of-compliance effect consistently shows that people substantially overestimate how likely someone is to say no to a direct request. The practical consequence is banal and uncomfortable: most of the requests you don’t make would have been accepted.
5. Make what you can do findable. Half of networking isn’t conversation: it is existing in a consultable form. A profile that states verifiably what you do works for you while you sleep. We cover it in optimising your LinkedIn profile.
6. Close the loop. If someone gives you useful information, tell them months later how it turned out. It is the single behaviour that turns a contact into a relationship, and almost nobody does it.
My take: a network isn’t something you switch on when needed
The first change I’d recommend isn’t meeting more people: it’s to stop treating your network as a resource to activate when needed.
People who only surface when they want something get polite, inconclusive answers, and I think the reason is exactly the one Casciaro describes: instrumentality is detectable, and it makes the receiver uncomfortable too. It hasn’t been measured, so it remains a hypothesis, not data.
Frequently asked questions
Is it true most jobs are never advertised? The 70–80% figure circulates without a verifiable source and deserves scepticism. What is documented is different and more useful: referred candidates are more likely to be hired, and weak ties causally increase job mobility.
I’m an introvert. Do I have to go to events? No. Nothing in the research cited here concerns industry events: it concerns network structure and reactivating existing contacts. One well-written message to someone who already knows you beats twenty handshakes.
How many contacts do I need? The right question isn’t how many, it is how different. According to Burt the value lies in the position between separated groups, not in the absolute number. Five different worlds beat fifty people from the same one.
What do I write to someone I haven’t spoken to in years? A concrete reference and a question that is easy to answer, with no request. If the conversation restarts, the request follows naturally; if it doesn’t, you wouldn’t have got it anyway.
Does networking replace applying normally? No, it runs alongside it. You still have to get through the interview, and on that we’ve collected the most common questions in job interview questions. If you’re considering a bigger move, the full path is in how to change jobs.
Sources
- M. S. Granovetter, The Strength of Weak Ties, American Journal of Sociology, 1973. Formulation of weak-tie theory and data on job referrals.
- M. S. Granovetter, Getting a Job: A Study of Contacts and Careers, 1974. Survey of the channels through which professionals found employment.
- S. Rajkumar, G. Saint-Jacques, I. Bojinov, E. Brynjolfsson, S. Aral, A causal test of the strength of weak ties, Science, 2022. Experiments on roughly 20 million users; moderately weak ties are the most effective for job mobility.
- S. V. Burks, B. Cowgill, M. Hoffman, M. Housman, The Value of Hiring through Employee Referrals, Quarterly Journal of Economics, 2015. Hiring probability, productivity and retention of referred candidates.
- R. S. Burt, Structural Holes and Good Ideas, American Journal of Sociology, 2004. Advantages associated with a brokerage position between groups.
- R. S. Burt, Structural Holes: The Social Structure of Competition, 1992. Original formulation of the structural hole concept.
- T. Casciaro, F. Gino, M. Kouchaki, The Contaminating Effects of Building Instrumental Ties: How Networking Can Make Us Feel Dirty, Administrative Science Quarterly, 2014. Moral discomfort associated with instrumental networking and the role of power.
- V. K. Bohns, (Mis)Understanding Our Influence Over Others: A Review of the Underestimation-of-Compliance Effect, Current Directions in Psychological Science, 2016. Systematic underestimation of how likely a direct request is to be accepted.



