A good café doesn’t know you from a profile. It knows you from Tuesdays. From the order you stopped having to say out loud around the third visit. From the nod at the door that means nothing to anyone else in the room and everything to the two of you. None of that was designed by anyone. It just happens, the way it’s been happening in cafés, pubs and corner stores for as long as those places have existed — and it turns out there’s a real body of research explaining why it works, and why the apps built to replace it keep struggling to.
The café was doing this before anyone called it a “third place”
In 1989, the sociologist Ray Oldenburg gave the phenomenon a name in his book The Great Good Place: the “third place,” distinct from home (first) and work (second) — the café, the pub, the barbershop, the bookstore, wherever people gather with no agenda beyond being around each other. Oldenburg’s criteria for a proper third place are almost defiantly low-tech: it’s neutral ground, nobody needs an invitation, it’s unstructured enough that you can come and go, and it isn’t expensive. Its whole job is conversation. The concept has become the standard reference point for why these spaces matter, and it’s been picked up everywhere from Brookings to UNESCO in the decades since.
Oldenburg also named something more specific: a third place gives its patrons “friends by the set” rather than friends one at a time. You don’t apply to belong to the 7am espresso crowd. You just turn up often enough, and one day you’re part of it — recognised, greeted, folded in without a form to fill out. It’s the regulars who make the room, Oldenburg argued, not the décor. They’re proof, just by being there again, that the place is worth returning to.
You don’t need a profile to be recognised — you need to keep showing up
There’s an older, stranger study that gets at the same idea from a different angle. In 1972, the psychologist Stanley Milgram — better known for the obedience experiments — ran what he called the “familiar stranger” study. He and his students photographed around 30 commuters on a New York subway platform, came back a week later, and asked other commuters which faces they recognised from their daily routine. Eighty-nine percent recognised at least one person, with an average of four familiar faces each — people they’d never spoken to, but had unconsciously catalogued through nothing more than repeated proximity. A 2004 replication by Intel’s Berkeley research lab found much the same thing: 77.8 percent recognised at least one familiar stranger, averaging 3.1 each.
Nobody swiped on those commuters. Nobody built them a match score. Recognition arrived purely through repetition — the same platform, the same time, often enough that a stranger’s face stopped being blank. It’s the same mechanism that turns a café into a place where the barista starts your order before you’ve said it. Presence is the whole algorithm. There’s no faster version of it, and no app has found one, because the thing being measured — a felt sense of “I know this person, roughly” — only accumulates in real time, in the same physical spot, over and over.
The people you barely know are doing more than you think
Here’s the part that cuts against how most people rank their relationships. In 1973, the Stanford sociologist Mark Granovetter published “The Strength of Weak Ties” in the American Journal of Sociology, one of the most cited papers in social science. Surveying how 282 people had found their jobs, Granovetter found that casual acquaintances — not close friends, not family — were disproportionately the ones who’d passed along the lead. Weak ties, he argued, connect you to information and networks your close circle simply doesn’t have, because your close circle mostly already knows what you know.
A café full of regulars is a weak-tie machine running on autopilot. The person two tables over whose name you don’t know but whose dog you do — that’s exactly the kind of loose connection Granovetter found actually moves things: jobs, introductions, a recommendation for a plumber, a passing comment that turns into something. Apps optimise hard for strong signal — the perfect match, the best-fit profile. Cafés, by just being a room people return to, generate the wide, low-stakes web of familiar faces that weak-ties research says does more useful work than anyone gives it credit for.
Compatibility isn’t something a profile can tell you
This is the part that should worry any product betting everything on the algorithm. In 2017, psychologists Samantha Joel, Paul Eastwick and Eli Finkel ran two speed-dating studies — 350 participants in total — where everyone completed more than 100 questions about their traits and preferences before meeting anyone, then went on a sequence of four-minute dates. Published in Psychological Science as “Is Romantic Desire Predictable?,” the study fed all that pre-date data into machine learning models and asked them to predict which specific pairs would click.
They couldn’t. The models predicted general desirability reasonably well — who’d be popular overall — but were unable to predict which particular two people would desire each other, no matter how the traits and preferences were combined. “We found we cannot anticipate how much individuals will uniquely desire each other in a speed-dating context with any meaningful level of accuracy,” lead author Samantha Joel told ScienceDaily at the time. Co-author Paul Eastwick put it more vividly: “Romantic desire may well be more like an earthquake, involving a dynamic and chaos-like process, than a chemical reaction involving the right combination of traits and preferences.” The chemistry, in other words, isn’t in the profile. It only shows up once two people are actually sitting across from each other.
What the café gets right
Put those four findings together and a pattern falls out. A third place works because it’s neutral, unstructured and cheap enough that showing up is easy. Recognition builds through repetition, not disclosure. The loose, half-known connections that form there do more social and practical work than the tight-knit ones. And whether two specific people actually get on isn’t something you can extract from a questionnaire beforehand — it’s only knowable in the room.
None of that is a knock on apps for trying. Matching, filtering and sorting are genuinely useful for narrowing a field of strangers down to size. But the part where a stranger becomes a familiar face, or an acquaintance becomes a lead worth following, or two people simply click — that part has never lived in the data. It lives in the fact of being physically present, on a repeat basis, without much else being asked of you. Cafés figured that out a long time before anyone wrote a line of matching code, mostly by not trying to solve it at all. They just kept the door open and the coffee coming, and let repetition do what no algorithm has managed to replicate.
Flat White starts from the same premise, just compressed: skip the months it takes to become a regular by accident, and get the one variable the research says actually matters — being in the room with someone — straight away. One café, one coffee, one hour.
Meet one person. Over coffee. On purpose.
Australia’s opening city-by-city. Join the waitlist at joinflatwhite.com →