Every few years an app appears promising to solve loneliness by connecting compatible people, and every few years the pattern repeats: plenty of sign-ups, a fair number of meetings, and very few that turn into anything. The usual conclusion is that it needed a better algorithm or more users.
Neither. The problem is not that matching works badly, it is that it solves a different problem. This guide sets out what these products actually optimise, why that is not enough for a bond to form, and what something genuinely aimed at loneliness would have to optimise instead.
What does a matching app actually optimise?
The first encounter. The whole machine — the profile, the algorithm, the interface — exists to maximise the chance that two strangers agree to meet once. And at that it works: it has never been easier to meet someone new.
The catch is that this is not the scarce step. Meeting someone you get on with was never the bottleneck of adult social life; the bottleneck is that most of those relationships never reach a second or third time, and the app takes no part in that. Its job ends precisely where the difficulty begins.
And there is a detail that is rarely made explicit: nothing in the business model rewards the relationship prospering. A product that charges for access to encounters has its metric in encounters, not bonds, and a relationship that works removes two users from the system. No bad faith needs to be assumed; it is enough to look at what gets measured.
Why doesn't a good match become a relationship?
Because a match delivers compatibility, and compatibility was almost never what was missing. Adult friendships do not fail from incompatibility; they fail from the absence of a recurring reason to meet again.
Think about how it goes: two people meet, it goes well, they part saying "we should do this again". From there, every subsequent meeting requires someone to propose, the other to reply, and both to find a slot. Each repetition costs the same as the first, and that constant cost competes with the whole of each person's life. Most relationships do not die of rejection: they die because nobody proposed the fourth one.
It is exactly the mechanism that makes adult friendship hard in general, and the app does not touch it. It has solved the pairing and left the expensive part untouched.
Which near-misses were already about to happen?
A great many, and that is the opportunity matching skips. In any city there are dozens of people daily who were going to do the same thing, nearby, at the same time, and missed each other by a tiny margin: two neighbours running the same route twenty minutes apart, two parents from the same school at the same park on alternate days, two people from the same building signed up to the same gym.
Those near-misses do not need a relationship invented from scratch. They need a very small push, because nearly everything else is already there: real proximity, a shared activity, an independent reason to be present. The marginal cost of turning them into an encounter is tiny next to arranging a date between two strangers with nothing to do together.
Classic matching ignores this whole layer because its unit of work is the compatible person, not the occasion that already existed. It searches the space of profiles when the abundance is in the space of overlaps.
What should it optimise instead?
Repetition, not pairing. And that changes the entire product, not the algorithm.
A system built for repetition looks little like a profile app. Its unit is not the person but the recurring occasion — a Tuesday five-a-side, a group that walks on Sundays — because an occasion that already repeats obliges nobody to propose anything. It does not need explicit reciprocity to work: joining something that was going to happen anyway is not asking anyone to give up an evening for you, so it never creates the running tab that kills most new relationships. It tolerates irregular attendance, because penalising absence turns a habit into an obligation. And its success metric is the fourth attendance, not the first: the first only proves interest, the fourth proves a habit has formed.
That is the useful test for any product claiming to fix loneliness, including any we build ourselves: ask what it measures. If it measures first encounters, it is solving a problem that was already solved.
Frequently asked questions
- So are apps for meeting people useless?
They do what they were built for, which is the first encounter, and that has real value for anyone who has just moved or changed their life completely. What they do not do is sustain what they start, and the mistake is expecting the second half from them. They are a doorway, not a place to stay.
- Why can't I be bothered to meet someone I liked?
Because proposing costs something and risks a no, while not proposing costs nothing today. It is an asymmetry that recurs every single time and usually wins. It is not indifference: the system demands a voluntary act at every iteration, and no new relationship survives that for long.
- Is it worth pushing through with more matches?
It is worth it for more first encounters, not more relationships. If the bottleneck is repetition, widening the intake only widens the top of the funnel: more people met, the same proportion never reaching a fourth meeting, and considerably more fatigue.
- How can I tell whether something will work before trying it?
Check whether it forces you to propose. If every subsequent meeting depends on someone taking the initiative, it reproduces the problem behind a different interface. If instead something already happens at a fixed time with or without you, you have the one piece that is genuinely scarce.