All articles ←

Three centimetres: on dry filters and what they cost

6 September 2026·7 min read·By The Zugot Project team

A person sets a height range, and from that moment the system hides everyone outside it — including a candidate who would have matched them on ninety-two percent of every other dimension. They will never know that candidate existed.

A dry filter

There are two kinds of criteria in any matchmaking system. There are criteria describing who you are — outlook, character, lifestyle, values — and there are criteria that are simply numbers or fields: age, height, region, head-covering style.

The second kind can be called dry filters. They're easy to collect and easy to compare, and so on every dating site they're given far more room than they deserve. A site that struggles to measure chemistry will measure height — not because height matters more, but because height is available.

The research on online dating points at exactly this: online systems assign excess weight to attributes that are convenient to search on, while what determines whether a relationship works is experiential attributes — humour, ease, conversational rhythm — that no field captures.

And yet, don't abandon the filters

The obvious conclusion is to abolish dry filters, and it's wrong. Someone who set an age range did so for a reason, and often an entirely substantive one — family planning, life stage, where they are in the process. A system that ignores the stated range isn't liberating the user; it's wasting their time and losing their trust.

So the filters stay strict. Anyone outside the range you set isn't offered to you. That's the default behaviour, and it applies in the overwhelming majority of cases.

The exception: once, and on conditions

There's one situation where the rule changes, and it's deliberately narrowly drawn. When the match on every other dimension — values, character, lifestyle, outlook — comes out extraordinary, above a high threshold, and the only deviation from the dry filters is a single one, by a small margin — the system doesn't reject quietly. It asks.

All three conditions are necessary, and each does work. The high threshold ensures nobody is bothered on behalf of a mediocre match. The single-deviation requirement prevents the case where a candidate misses on age and on height and on region, each "small" on its own while the combination no longer is. And the small margin — two years on age, five centimetres on height, a neighbouring region rather than a distant one — defines what counts as a slight miss.

And even then, every other hard filter continues to apply in full. An extraordinary match opens nothing except that single clause that was missed by a little.

Why ask rather than decide

The significant design decision here is that the system doesn't widen the range by itself. It presents the case to the user and asks for approval — telling them exactly what the deviation is and what the score is.

The difference isn't semantic. A system that quietly widens ranges trains its users not to trust their own settings: if I asked for up to thirty and I'm getting thirty-two-year-olds, I conclude the numbers don't really work, and next time I'll set a narrower range than I actually want to compensate. That draws both sides into a game.

An explicit question does the opposite. It leaves the setting valid and hands the exceptional decision to the person it belongs to. And if the answer is no — that's a completely legitimate answer, and the candidate doesn't come back.

What this means when you set ranges

A few practical conclusions. First, set the ranges as they genuinely are, not narrower "to be safe". A too-narrow range doesn't improve candidate quality — it only reduces the number, and mostly it cuts off the tail where the unusually strong matches sit.

Second, when such a question arrives — it's worth reading the score in it. A message saying the match on character and values is very high, and that the only gap is three centimetres, is giving you real information. You can say no, but it's worth saying no knowingly rather than out of habit.

And third, it's worth asking where the number came from. A great many age and height ranges are written because a field needed filling, not because they were examined. A number you never tested filters people out just as efficiently as one you weighed carefully.