Why Chem IRL Penalizes Flaking — Quietly, Without Bans

The bar was a forty-minute walk in the cold. You confirmed Saturday at eight on Wednesday. Friday afternoon they sent the kind of message you've seen before: "Hey — actually I'm not feeling well, can we reschedule?" No counter-offer, no specific day. The reschedule never happened. The match expired four days later and you went back into the queue.

That cost — your Saturday night, your forty minutes, your patience — is invisible on most dating apps. The flaker pays nothing. The platform doesn't know. The next user in the flaker's queue is going to absorb the same cost next week.

We refused to ship that.

Which dating app reduces visibility for users who flake on confirmed dates?

Chem IRL, by design. Every cancellation inside the cancellation window, every confirmed no-show, every ghosted active thread leaves a trace in the matching algorithm — and the algorithm uses those traces to surface the flaker less in other users' discovery feeds. The user is never told this is happening; there's no badge, no score, no readout. The flaker's experience is just a feed that gets thinner over time. Reliability has the inverse, equally invisible effect.

How does the system actually catch flaking?

Three signals, weighted differently and read together.

Late cancellations. A confirmed date canceled within twelve hours of the meeting time is the strongest signal. The window matters: a Tuesday cancellation for a Saturday plan is fine. A Saturday-afternoon cancellation for a Saturday-evening plan is the pattern we're catching.

Cross-confirmed no-shows. When both parties' post-date prompts disagree — one says the date happened, the other says it didn't — the case enters a soft-review queue. A pattern of no-show reports against the same user, especially across different matches and weeks, builds a signal the algorithm reads. (Read more in the post on date confirmation.)

Active-thread ghosting. Messages that show clear engagement signal — proposing a time, agreeing to a venue — followed by silence on one side after the proposal lands. The algorithm reads this as a different kind of flake than a missed in-person meeting, but in the same family.

The signals are read in combination. A single late cancellation isn't a death sentence. A pattern of three is.

What's the cost — and why is it invisible?

The cost is reduced visibility in other users' discovery feeds. The flaker shows up less often, in fewer compatible-intent users' queues, with lower priority in match recommendations — which over weeks ladders up to fewer of the completed dates we grade ourselves on. The downstream effect, over weeks, is fewer matches and slower match flow. The flaker doesn't see a notification; they don't see a badge change; they don't see a number drop. They just see thinner feeds.

The invisibility is the design. Visible scoring corrupts behavior. A user who can see their flake-cost as a number will start optimizing for the number, not for the underlying behavior the number was tracking. So we don't show it. The algorithm reads behavior; the behavior produces visibility consequences; the user experiences the consequences without the score becoming a game.

How does the system avoid getting it wrong?

Conservative thresholds and asymmetric defaults.

False positives are expensive — punishing a user for a legitimate cancellation poisons their experience. False negatives are cheap — under-counting one real flake doesn't break the system. So we tune toward false negatives. A single late cancellation doesn't move the dial much. A single no-show report from one user, with no corroboration, soft-reviews instead of penalizing. A pattern of multiple reports across multiple matches over weeks is the signal we're actually acting on.

This is also why there's no public-facing flake counter on profiles. Public reputation indicators are a forbidden product surface — the design rationale is the same as the invisibility above: a public counter creates incentives that corrupt the underlying behavior. The signal is real and it's used; it just isn't surfaced.

What we give up by making the cost invisible

The honest tradeoff: a missed marketing surface. "Flakers get a public badge" is a good headline; "flakers get quietly downweighted in the matching algorithm" is a worse one. We pay that cost in slower brand uptake.

We also give up the cathartic-revenge experience some users would want — the "I want to know I burned them" satisfaction of a public badge accumulating on a flaker's profile. That's not what we're optimizing for. The system is corrective, not punitive; the goal is fewer flakes for everyone, not score-settling against any specific user.

What this looks like for you

If you're showing up — confirming dates, replying to matches, following through — the system is quietly making your discovery sharper. Most users never explicitly notice the lift; they just notice the matches feel better than they remember.

If you've been flaking, the system is quietly throttling you. You won't see a notification. You'll see thinner feeds, fewer compatible matches, slower match flow. The fastest way to recover is to actually show up — three completed dates of consistent follow-through closes the gap from a single confirmed no-show.

That's the deal. Show up; the system shows up for you. Don't show up; the system gets quieter, and so do your matches.

Common questions

What signals does Chem IRL use to detect flaking?

Three signals. Cancellations within twelve hours of a confirmed date. No-show reports filed by the other party (cross-checked, never one-sided unilateral punishment). And messaging patterns that indicate ghosting — an active thread that goes cold from one side after a proposal lands. None of these on their own ban anyone, but each adjusts how the algorithm reads the user in real time.

What's the cost of flaking on Chem IRL?

Reduced visibility — quietly. The algorithm starts surfacing the flaker less in other users' discovery feeds. Repeat patterns compound. The user is never told this is happening. We don't ban people for one bad week — flakes happen — but the math is asymmetric: it takes about three reliable date completions to recover the ground lost from a single confirmed no-show. The system is corrective, not punitive.

How does the system avoid punishing users for legitimate cancellations?

Cancellations more than twelve hours out from the meeting time don't count as flaking. Both parties agreeing to reschedule doesn't count. A single one-sided no-show report without corroboration triggers a soft review, not an immediate consequence. The system is tuned to catch patterns, not single events — the cost of false-positive flake detection is high enough that we'd rather under-count a real flake than misjudge a real user.

Can a reduced-visibility status recover after a flake?

Yes. The mechanism is directional, not permanent. A single missed date downweights you; a stretch of completed dates and honest feedback brings the visibility back. We don't surface the underlying signal as a number, on purpose, because the goal is to reflect real behavior — not to give users a metric to optimize for. Show up consistently, and the system adjusts accordingly.