A single number for the quality of a room.
SignalScore™ measures relevant, reciprocated, followed-through connections. Computed from what actually happened in the room, never from headcount, badge scans or app opens.
Two connections. Both mutual, both relevant, both replied to. That is an 86. Depth beats breadth.
Most tools measure volume. Volume is easy to game and tells you nothing.
Tickets sold, badges scanned, messages sent. None of it says whether the room worked, whether people left with connections that matter. SignalScore™ is built on four principles.
Quality over volume
The system caps how many people one attendee can reach. You cannot buy a high score by working the whole room.
Reciprocity is the signal
A one-way save is noise. A mutual tie, where both saved each other or both replied, is the atomic unit of a real connection.
Nobody leaves alone
The heaviest weight of all is inclusion: the fraction of the room that left with at least one real tie. Not a place where the popular get more popular.
Always explainable
Every score ships with its sub-scores. Never a black box. Personal scores stay private, never a leaderboard.
An edge is interest. A tie is a connection.
Tap any link in the room to cycle it: nothing, a one-way reach (a save or an intro), then a reciprocated tie (both saved, or both replied). Only ties move the score.
Only reciprocated ties move the number. One-way reach is just interest.
Any sign of interest: a save, a profile view, an intro. One direction only.
A reciprocated edge. Both people saved each other, or both replied. This is a real connection.
How much of the interest in the room turned mutual instead of staying one-way.
The share of attendees who left with at least one real tie. The thing that matters most.
How well two people actually fit: what one offers against what the other needs.
How far a connection travelled: from just showing up, to a two-way conversation.
How well did I network? Four inputs, one honest number.
Move the sliders and watch a live model of the score respond. It shows what each input does. The exact weighting stays under the hood.
Mutual connections you actually built.
Everyone you saved or messaged. Reciprocity quality is ties over reach.
AI-scored fit between what you offer and what they need.
The furthest stage you reached. Set it in the funnel below, too.
Dana at a 40-person founder dinner. She reaches out to six people over the night. Three of them save her back or reply: three real ties from six attempts, so her reciprocity quality is high. Two of those ties are a genuine fit for what she's building, and she gets a reply from one before she's even left. Modest reach, but deep and mutual: Dana walks away with a strong personal score, and the app can tell her exactly which of the four inputs earned it. The person who added forty contacts and heard back from none scores far lower.
A ratchet, not a sum. You get credit for the furthest point you reached.
Tap a stage. It mirrors a real relationship funnel: showing up, noticing someone, saving them, reaching out, then a two-way conversation. This feeds follow-through in the score above.
How well did the room connect? Five rates, comparable across any size.
Inclusion carries the heaviest weight. The dashboard always surfaces the weakest link, so an organizer knows exactly what to fix next.
Fraction of the room who made at least one mutual tie.
Of pairs who interacted at all, the fraction that went mutual.
Average fit of the mutual ties that formed.
Largest connected cluster over the whole room. One web, or scattered cliques.
Of conversations started, the fraction that got a reply.
If relevance is unavailable, its share is redistributed, not dropped.
Turn the AI relevance signal off. Its share is spread proportionally across the other three components, so they still sum to one and the score stays on the same 0–100 scale.
Four components share the score. With the AI signal on, relevance takes its own slice of the ring. The same graceful degradation protects every AI input: if the model is off, the system falls back to a deterministic heuristic with zero API calls, and the rest of the score runs unchanged.
A metric you can defend, line by line, in front of a board.
Deterministic, unit-tested, and tuned from a single reviewed file. Every guardrail exists so the number holds up under scrutiny.
Weights sum to one
Enforced at startup. Break the sum and the app refuses to run. No silent miscalibration.
Volume is capped
Strict per-person limits on how many people anyone can reach. The score can't be brute-forced by working the room.
Pure and unit-tested
The scoring core does zero input or output. Hand-calculated fixtures verify every result.
One tuning surface
Every weight and threshold lives in one reviewed config. Re-tuning is a single, auditable change.
The score becomes action. Every attendee gets a segment.
Classified by the room's own distribution, so thresholds scale with size and never hard-code a popularity bar. This is what tells an organizer who to seat next to whom, and who nearly slipped through.
Hub
A natural connector, many mutual ties.
Broker
Bridges otherwise separate clusters.
Specialist
Few ties, but high relevance. Quiet value.
Wallflower
Present, no tie yet. The person to help next.
Connected
A healthy, ordinary middle.
Personal scores are for one person only. Never a public ranking.
Enforced at the database with row-level security. Every table is scoped per organization, all computation runs server-side, and attendees never receive anyone else's raw activity. Explainable to the person it belongs to, invisible to everyone else.
A cross-event, human-verified layer: trust, opportunity, and real outcomes.
The scores above read a single event from behavior. This layer adds what behavior alone can't see: off-platform follow-up, real outcomes, and trust built over repeated events. It also tracks a longitudinal trend that shows whether your events are getting better at connecting people over time.
Trust score
A blend of five repeated signals built over time: familiarity, similarity, competence, reciprocity and reliability.
Business opportunity
Trust times relevance times timing. Multiplicative on purpose: if any factor is zero, business doesn't happen.
Post-event survey
Two automated waves capture real outcomes and off-platform follow-up. Non-response never penalizes anyone.
The questions people ask first.
Can someone game their way to a high score?
It's built to resist exactly that. Volume is capped per person, and only reciprocated ties move the number. Blasting the whole room with one-way saves does almost nothing. Depth beats breadth by design.
Is my personal score private?
Always. A personal SignalScore™ is shown to one person only, never as a public ranking. It's enforced at the database with row-level security, scoped per organization, and computed server-side. Attendees never see anyone else's raw activity.
Do you publish the exact weights?
No. Every score is fully explainable through its sub-scores, so you always know why it moved. But we keep the precise weighting private, so it can't be gamed or copied.
What if the AI relevance signal isn't available?
The score never breaks. Relevance's share is redistributed across the remaining components, which are rescaled so everything still sums to one and stays on the same 0–100 scale. The system falls back to a deterministic heuristic with zero API calls.
Why doesn't a tiny event get a score?
Below a minimum number of check-ins, SignalScore™ returns no number at all. A very small room produces a misleading figure, so we show nothing rather than something fake.
Measure what a room is really worth.
See SignalScore™ run on your own events, with a report you can hand to a sponsor or a board.