The Friendships You Lose Are the Ones Nobody Else Knows About

There is a name you have not said out loud in about two years. You liked them. There was no argument and no falling out, nothing you could point at as a reason. You got busy, they got busy, the thread went quiet in a way that felt temporary for roughly four months, and then it stopped feeling like anything at all.

Now think about a friend you have kept. There is a decent chance somebody else is in the picture. A partner who knows them. A sibling who has met them. Another friend who would text them if you went quiet. The friendship you kept has other people in it. The one that faded was private.

That is not a statement about how much you cared. It is closer to arithmetic. A friendship that exists only between two people has exactly one point of failure: either of you getting busy. Add one person who knows you both and the thing has a second reason to keep existing, one that does not depend on either of you remembering.

Most advice for a fading friendship is emotional. Feel more. Try harder. Be more honest about missing them. The network research points somewhere less romantic and a good deal easier to act on: stop trying to feel closer, and make your friendships less private.

1. The friendship with one point of failure

The most direct evidence on how relationships wear out comes from a study that tracked four annual surveys of working relationships among 345 people at a large financial firm. The headline number is bleak. Of the relationships at risk of ending in a given year, three in four were gone by the next one.

The interesting part is which ones survived. Decay was slower between people with a strong prior relationship, people in the same division, and – this is the one that matters here – people connected indirectly through many third parties. The paper goes further: it concludes that the greater stability of older relationships was carried by that indirect connection rather than by age on its own. Time was not what made the old ties durable. Other people were.

Worth being honest about the limits, and they are large. Those were colleagues at one bank, not friends. Work relationships are propped up by payroll and project assignments in ways friendships are not, and in that same paper colleague ties decayed faster than other kinds, not slower. Nobody has run a four-year design like it on adult friendships.

There is one smaller study closer to the target. A short paper re-examined a survey of first-year men at a small liberal arts college, measured in November and again in April, and defined the embeddedness of a tie as the overlap between two people’s social relations. It found that overlap is relatively stable, and that it tends to have some effect on whether relationships continue. “Some effect” is the author’s own hedge, and we are keeping it rather than upgrading it.

The move: the audit. Write down the five or six friendships you would be genuinely sorry to lose. Next to each name, write the name of one other person in your life who also knows them. Not someone who could be introduced to them. Someone who actually knows them. The blanks are your one-point-of-failure friendships. This takes four minutes and it is slightly uncomfortable, which is roughly the point.

2. Overlap is not proof of closeness. It is what holds closeness up.

The foundational paper on tie strength, published in 1973, makes a proposition that sounds obvious until you turn it around: the overlap between two people’s friendship networks varies directly with how strong the tie between them is. The stronger the friendship, the more names the two of you have in common.

Almost everyone reads that as a symptom. Close friends end up knowing each other’s people, the way close friends end up knowing each other’s coffee order. The network reading runs the other way as well. The overlap is not only what closeness produces, it is part of what keeps closeness alive, because a friendship with shared names in it is maintained by more than the two of you paying attention.

The college study above adds the mechanism, and it is the least sentimental line in this whole post: embeddedness is less under an individual’s control than other aspects of a relationship, which is precisely why it is more stable. A private friendship survives on effort. An overlapping one survives partly on other people’s calendars.

Be careful how much weight that carries. The 1973 proposition is theoretical and observational, not a test, and the college finding is a short paper on one cohort at one school. Neither demonstrates that adding a mutual friend to an existing friendship rescues it. What they establish is narrower and still useful: overlap and durability travel together, and overlap is the part you can change deliberately.

The move: the three-way thread. Take one name from your audit that had a blank beside it. Think of one person from another part of your life who would genuinely get on with them, and start a three-person thread with something concrete in it – a date, a place, a thing that is actually happening. Not a group chat for its own sake. One shared plan, three people, so the friendship now has a second person keeping it in view.

3. Making the introduction is a specific act, and it has been studied

There are two ways to sit between two people who do not know each other. You can keep them apart, which is the brokerage position network research is most famous for, where the advantage comes from being the only link. Or you can join them, and a 2005 study gave that second orientation a name and a measure: the third who joins, someone who habitually introduces disconnected people or helps already-connected people coordinate something new.

In that study, the joining orientation predicted involvement in innovation, alongside dense networks and diverse knowledge. Here is where we have to be straight with you. The setting was the engineering division of a car manufacturer, the outcome measured was innovation at work, and nothing in the paper is about friendship. What it does establish is that introduction-making is a real, measurable habit that varies a lot between people, rather than a vague social virtue.

For whether a mutual connection actually causes a new tie to form, there is much fresher evidence. A 2025 field experiment on a social platform varied at random whether a user already had a tie to an account that had followed the target user, then watched whether the user followed back. The minimal cue that a mutual connection existed raised tie formation by 35 percent, and the effect was larger when the user had interacted more with that existing connection. Strong shared ties introduce better than weak ones.

Two limits the authors put in print themselves, and we will not bury them. That 35 percent is substantially smaller than what the older observational studies imply, which is a warning about most confident claims you have read on this subject. And a follow on a platform is not a friendship. It is the cheapest possible version of saying yes to someone.

The move: the introduction with a reason. Send one message this week that puts two people together, and make it specific. Not “you two should meet.” Name the overlap in one line – what each of them is actually doing right now, and why that is interesting to the other – and then get out of the way. Three sentences, under ten minutes. The reason is the whole thing, because an introduction without one hands the work of finding the point to two strangers.

4. People get into a group through the people already in it

If your honest problem is not one fading friendship but a general thinness, the research on how people join things is unsentimental about the route. A 1992 study modelled how voluntary groups gain and lose members and found that the probability of entering or leaving a group depends on the number and strength of the network ties connecting members to each other and to nonmembers. Ties inside a group were associated with staying. Ties reaching outside it were associated with leaving, and with joining somewhere new. Its conclusion is that those outward-facing ties are a major source of change in who belongs where.

Where new relationships actually appear has been tracked too. A Dutch panel followed personal networks over seven years and found that a lack of meeting opportunities is an important reason relationships get discontinued, and that a new relationship is more likely to emerge in a setting where you already meet people you know. New ties mostly form in rooms you are already in, beside people you already have.

The limits, stated plainly: the 1992 work is about memberships in voluntary organisations, using survey data from a different decade, and a membership is not a friendship. The Dutch study asked about confidants and practical helpers, and its authors specifically caution against over-reading network change measured with a single name-generating question. Neither is a controlled test of getting invited to things.

The move: the guest route. Pick one person you already know and ask what they do regularly – the run, the class, the Tuesday thing, the table they sit at. Then ask to come once. You are not looking for a new group, you are using an existing tie as the door, which is how the research says people actually get inside one. Going once with somebody beats going alone to something new.

The through-line

The friendships you lose are rarely the ones you cared about least. They are the ones that existed nowhere except between the two of you, so a single busy season was enough to end them. None of the repairs asks you to feel anything you are not already feeling:

• Write one shared name next to each friend you would hate to lose. The blanks are the fragile ones.
• Take one private friendship and put a third person into something concrete.
• Make one introduction this week, and write the reason into the message.
• Ask one friend what they do regularly, and ask to come along once.

You cannot make yourself less busy. You can make a friendship that fewer things have to go right for.

Sources

• Burt, R. S. (2000). Decay functions. *Social Networks*, 22(1), 1-28. https://doi.org/10.1016/S0378-8733(99)00015-5
• Feld, S. L. (1997). Structural embeddedness and stability of interpersonal relations. *Social Networks*, 19(1), 91-95. https://doi.org/10.1016/S0378-8733(96)00293-6
• Granovetter, M. S. (1973). The strength of weak ties. *American Journal of Sociology*, 78(6), 1360-1380. https://doi.org/10.1086/225469
• Feld, S. L. (1981). The focused organization of social ties. *American Journal of Sociology*, 86(5), 1015-1035. https://doi.org/10.1086/227352
• Obstfeld, D. (2005). Social networks, the tertius iungens orientation, and involvement in innovation. *Administrative Science Quarterly*, 50(1), 100-130. https://doi.org/10.2189/asqu.2005.50.1.100
• Mosleh, M., Eckles, D., & Rand, D. G. (2025). Tendencies toward triadic closure: Field experimental evidence. *Proceedings of the National Academy of Sciences*, 122(27), e2404590122. https://doi.org/10.1073/pnas.2404590122
• McPherson, J. M., Popielarz, P. A., & Drobnic, S. (1992). Social networks and organizational dynamics. *American Sociological Review*, 57(2), 153-170. https://doi.org/10.2307/2096202
• Mollenhorst, G., Volker, B., & Flap, H. (2014). Changes in personal relationships: How social contexts affect the emergence and discontinuation of relationships. *Social Networks*, 37, 65-80. https://doi.org/10.1016/j.socnet.2013.12.003