The loneliest version of a day is rarely an empty one. It is a full calendar, a busy workplace, messages arriving on schedule, and a ride home surrounded by people, with none of it reaching the part that actually needs company. Most products built for connection optimise for more contact. Building a peer support site kept pointing the other way: the problem was not the amount of contact, it was that almost none of it allowed the unedited version of a day to be said out loud.
That distinction changed the design. If the real gap is between being busy and being known, then adding another feed, another follower count, or another group chat solves nothing. The work became matching: putting a person in a small circle of people who have lived the same stretch of life, so the explanation tax at the start of a conversation drops close to zero.
Designing for the unedited version
The first choice was privacy by default. Every post starts private to the circle, not public to the platform. That sounds like a growth handicap, and in raw signup terms it probably is. But the earliest test conversations showed a clear pattern: people wrote the honest version only when they believed it would stay inside a small, defined room. The edited version is what every other app already gets.
The second choice was matching on lived experience instead of interests. Two people who both like hiking have a topic. Two people who have both sat through a first winter after a loss, or a first year of caring for a parent, have recognition. The intake therefore asks what someone is going through, in plain words, and treats hobbies and demographics as decoration rather than signal.
The third choice was circle size. Small enough that a name becomes familiar within a week. A circle is not an audience, and the interface work kept removing audience mechanics: no public counts on posts, no ranking, nothing that rewards performing a recovery story over describing an ordinary Tuesday.
What worked
One ordinary detail beat the big opening statement. The first prompt originally asked people to share their story, and most drafts died there. Replacing it with a request for one plain detail from this week, a meal, a phone call, an errand that went sideways, got first posts written. Recognition usually starts small.
Matching on the season, not the event label, worked better than expected. Two people rarely describe the same experience with the same words, so exact label matching produced near misses that felt worse than no match. Grouping by the shape of the season, new and acute, long and grinding, or slowly improving, put people together whose days actually rhymed.
Private by default increased reply depth. Threads inside circles run longer and more specific than the public posts the site launched with, and people return to the same thread across weeks. The honest limit is that depth is hard to show in a screenshot. It shows up in return visits instead.
What broke, or nearly did
Topic tags were the first matching system, and they failed quietly. Circles formed around a shared tag often had nothing else in common, conversation stalled after introductions, and the stall itself felt like rejection. The tag system was not wrong as a filing tool. It was wrong as the reason people were put in a room together.
Intake questions kept turning into forms. Every sensible question added to help matching also raised the cost of starting, and the completion data punished each addition. The fix was cutting the intake to the few answers that actually change a match, and letting the rest be learned from what a person chooses to post first.
Bigger circles felt safer to build and worse to be in. Doubling circle size raised the chance someone would reply quickly, and lowered the chance anyone wrote honestly. The room started behaving like a feed, with the most polished posts getting the attention. Shrinking back down lost speed and kept the trust, which is the trade the product actually runs on.
Lessons for builders
Count the explanation tax. If a new user has to explain their whole situation before anyone can respond usefully, the matching has failed before the conversation starts. Design so the room already knows the shape of the thing.
Privacy is a feature people can feel in the first minute. Defaults decide what gets said. A private default does not just protect people, it changes the quality of what they are willing to write, and that quality is the product.
Loneliness is not solved by contact volume. It is eased by recognition, and recognition is a matching problem with a specific, buildable shape: the right few people, in a private room, at the same season of a hard thing.
The project behind these lessons is SupportMeet, a free peer support site where people share what they are going through and are matched with circles who have lived the same thing, private by default, with help in every country.
