What is a signal?

A woman signs up for her company's spring conference. That is a fact: a name on an attendee list, true and almost meaningless on its own. Thousands of people do the same.

Then she chooses the breakout session on rebuilding trust after a reorganisation, and so does a man from a team three floors away who is about to lead exactly that. Now the fact has grown into something more interesting: not proof that they should meet, but a hint that a conversation between them might be worth having. Add that both of them, when they joined, quietly ticked a box saying they were open to meeting peers, and the hint grows stronger still.

None of this has decided anything. Neither of them knows the other exists. But somewhere in the plain record of what they each did, a reason to introduce them has assembled itself. That reason, and the care it takes to read one honestly, is what this article is about. It has a name: a signal.

A working definition

A relationship signal is an appropriate piece of context that suggests two people may have a good reason to connect.

Every word in that sentence is load-bearing. Appropriate, because not everything that could be known should be used. Context, because a signal is about a situation, not a person's character. Suggests, because a signal proposes rather than concludes. And may, because it deals in possibility, never certainty. A signal opens a question. It does not answer it.

It helps to hold four things apart that are easy to blur: evidence, interpretation, opportunity, and decision. A signal is interpretation laid over evidence. It can become an opportunity worth surfacing. But the decision, whether anyone actually meets, belongs to the two people, and to no signal, however strong.

A fact is not yet a signal

The most common mistake is to treat every fact as if it were already a recommendation. It is not. Facts, events, and actions are only the raw material; they become signals when context gives them meaning. Watch a plain fact turn, step by step, into something worth acting on:

Read it the other way and the discipline is clear. Attending an event is a fact. Both attending the same narrow session may become a signal. Both having said they would welcome meeting peers strengthens it. Being free at the same hour makes the resulting opportunity actionable. And still, at that point, neither person has agreed to anything. The evidence has done its work; the choosing has not yet begun.

Where good signals come from

If facts become signals only in context, the honest question is which facts a system should look at in the first place. Not everything that can be observed should be. Responsible signals tend to come from a few well-lit places.

The strongest and safest are declared: things people intentionally chose to share. What they are interested in, what they want to explore, the support they can offer or are seeking, whether they are open to mentoring, how and when they prefer to meet, the language they are most comfortable in, the access they need. Declared information is trustworthy precisely because the person meant to provide it. A good system leans here first.

Membership and shared spaces are a second source: the same organisation, event, cohort, school, association, team, or cross-functional project. On its own, shared membership is weak. Ten thousand people work at the same company, and that fact alone recommends none of them to each other. Membership matters only when narrower context makes it matter.

Events and participation can carry more meaning: attending related sessions, joining the same programme, contributing to a shared initiative, volunteering for the same cause, taking part in a discussion. The caution here is real. A system should read only the participation that is expected and purpose-bound, the ordinary visible marks of taking part, and never drift into monitoring everything a person does simply because it can.

Transitions are among the richest signals of all, because they open temporary windows. Joining a new organisation, moving to a new city, stepping into a new role, returning from leave, starting a programme, finishing onboarding, preparing for a career change: each makes certain conversations briefly and genuinely relevant, in a way that will fade. Someone who navigated last year's move is exactly the right person to meet this month, and beside the point a year from now.

Offers and requests are the signals that fit together like a key and a lock: one person asks for help with a topic, another has offered experience in it; a mentor is open to newcomers, a newcomer is looking for one; an attendee wants to continue a specific conversation, and so does someone else. When a stated need meets a stated willingness, the reason to introduce almost writes itself.

And timing and availability quietly govern all of it. A theoretically perfect introduction is a poor one if someone is overloaded, has paused introductions, is no longer pursuing the objective, or simply is not free. Timing can strengthen a signal into an opportunity or, just as easily, dissolve it. The same two people can be a good idea in March and the wrong idea in April.

Signals have strength, not scores

Signals are not all equal, and pretending they are is how systems become both shallow and unsettling. They vary in strength, and strength usually comes from several honest indicators pointing the same way rather than one doing all the work.

You can feel the difference building:

• A weak signal: both people list 'innovation' as an interest. So do half the organisation.
• Stronger: both attended the same workshop on innovation in the public sector.
• Stronger still: one asked for perspective on actually implementing it, and the other has done exactly that.
• An actionable opportunity: both are open to a short introduction, and their calendars can find a half-hour.

That is a real progression, and it is worth being precise about what it is not. It is not a percentage. The moment strength is dressed up as '87% compatibility' or a 'connection probability' or a 'relationship score', it has stopped being honest, because it claims a precision about two human beings that no evidence can support. If a system tracks confidence at all, that confidence should describe one thing only: the quality of the evidence behind a particular opportunity. It is a statement about how good the reason is, never a rating of how good the people are.

Signals read best together

A single signal is rarely enough to act on, and a pile of the same signal is not much better. Signals earn their meaning in combination, and the most useful combinations often mix sameness with difference:

• shared context plus complementary experience: you were in the same room, and you each know something the other does not;
• a current objective plus offered support: one person needs the very thing another has volunteered to give;
• event participation plus mutual openness: you both showed up, and you both said you were glad to meet people;
• repeated proximity plus no prior introduction: you keep ending up in the same places and have somehow never actually met;
• a shared transition plus relevant lived experience: you are going through what the other person came through last year.

Notice how often difference does the real work. Two people who are too alike may have little to offer each other beyond agreement; the conversations that teach us something usually cross a gap. Good signal-reading looks for enough common ground to make a meeting easy, and enough difference to make it worth having.

Signals that say wait, or no

It is tempting to think of signals only as green lights, but the honest ones point both ways. Some of the most important signals are the ones that say this is not the moment, or not the pairing:

• someone has paused introductions;
• the suggestion runs against a privacy preference they set;
• one person has already declined this kind of meeting;
• the timing is plainly bad for one of them;
• there is a reporting line or a power gap that makes a casual coffee awkward;
• the context is too sensitive to be handled this lightly;
• the only available reason is thin;
• these two were introduced very recently;
• the community's own rules do not allow the match;
• someone has reached a sensible limit on how much they want to meet.

The right way to read these is as constraints and missing conditions, not judgements about anyone. A person who has paused introductions is not a 'bad match'; they are simply, reasonably, not available right now. A thin reason is not a flawed person; it is just not yet enough reason. A careful system treats a limiting signal with the same respect it gives an encouraging one: as information about a situation, never a verdict on a human being.

Signals are not surveillance

All of this depends on a line that must never be crossed, and it is worth drawing in bright ink. Reading signals is not the same as watching people, and a system that cannot tell the difference should not be trusted with either.

A responsible signal system does not read private messages, record conversations without consent, or infer sensitive traits nobody chose to share. It does not monitor private behaviour, track people beyond the purpose of the community they joined, build hidden personality profiles, or hoard data simply because it is technically available. It does not repurpose workplace systems for unrelated social ends. And it does not treat silence, delay, or non-participation as if they revealed something about a person's character, because they do not.

What it does instead can be said in plain principles: use information only for the purpose it was given, collect as little as possible, be open about what is used and why, honour the boundaries of the space a signal came from, respect the preferences people set, keep nothing longer than it is useful, be able to explain any introduction in words, and let people correct or withdraw their information at any time. None of that is exotic. It is simply the difference between understanding people and spying on them.

Where a signal came from matters

Not all signals deserve the same trust, and one of the best guides to how much weight a signal should carry is its provenance: where it came from. A system should always know, and treat differently, whether a signal is:

• user-declared, offered deliberately by the person themselves;
• organiser-provided, supplied by a community or event with its own purpose;
• observed participation, the visible marks of taking part;
• inferred, guessed at from other information rather than stated;
• drawn from earlier feedback, how a previous introduction actually went;
• external, from beyond the community, if it is ever permitted at all.

The rule of thumb is strict: the more inferred or sensitive a signal is, the more restraint and human review it demands, and the less it should ever stand on its own. A declared preference can be trusted at face value. An inference is a guess wearing a confident expression, and guesses about people should be handled with humility, checked against what the person actually chose to share, and never allowed to quietly override it.

A reason you can read

Everything a signal is for comes down to one test the person on the receiving end can apply in a second: does the introduction come with a reason I can actually understand? A signal that cannot be turned into a plain-language reason is not ready to be acted on.

Good reasons name the shared context and the willingness:

• You both recently joined the same organisation and said you'd like to meet peers.
• You attended related sessions and are exploring the same question.
• One of you offered to help with onboarding, and the other is going through it now.
• You work in different teams on connected problems, and you're both open to a cross-team chat.

Bad reasons hide the evidence behind a machine:

• Our model selected you.
• You have a high compatibility score.
• The algorithm found a strong connection.
• Your behavioural patterns are aligned.

The first kind hands the reason to the person and lets them judge it. The second asks them to trust the system's opinion of them. Only the first is compatible with treating an introduction as an offer rather than an instruction. This is the same point at the heart of the argument that profiles are not enough: a good signal, like a good introduction, can always say why.

Reading a signal well

If you ever want to judge a proposed introduction, yours or a system's, a handful of questions do most of the work:

• What actually happened, in plain terms?
• Why might it matter to these two people?
• Is the context still current, or has the moment passed?
• Is the information being used appropriate to use at all?
• Can the reason be explained in a sentence?
• Does the potential value run both ways?
• Are both people open to this kind of introduction?
• Is there enough here to justify surfacing it, or is it still just a fact?
• And is there any reason not to proceed?

That last question matters as much as the rest. Good signal-reading is not only about finding reasons to connect people. It is about being willing to see the reasons not to.

A reason, gathered with care

Do all of this well, honestly, with restraint, in context, and you have accomplished exactly one thing: a good reason to believe two people might be glad to meet. But notice how much of the care was really about restraint: which facts to use and which to leave alone, how close to stay to the evidence, when to say no.

That restraint is not incidental, and it raises a question that deserves its own answer. If a system leans on context to understand who might connect, how does it do that without sliding into watching people? A relationship signal and an act of surveillance can look alike from a distance, and everything depends on telling them apart. Relationship intelligence without surveillance is where the journey goes next.