Algorithm guide
How matching works
A plain-language guide to who you see and why. It's a published, deterministic baseline — no hidden pools, no paid exposure.
Deterministic baseline: hard mutual filters first, then a weighted score of reciprocal intent, shared interests, activity timing, proximity and freshness, with a cold-start fairness slot and no popularity or attractiveness signals.
1 · Hard mutual filters come first
Before anything is ranked, every candidate must pass these checks in both directions. If it doesn't fit for either person, they're never shown — no score can override this.
Adults only
Both people are 18 or older, with active, non-restricted accounts.
Mutual interest
For dating, each person is someone the other wants to meet.
Mutual age ranges
Each person's age fits the other's chosen range.
Mutual distance
Within both people's maximum distance (measured between snapped points, never exact positions).
No blocks
Blocks hide people from each other everywhere — discovery, plans and chats.
Passes are respected
People you passed don't come back unless you reset recommendations yourself.
Dealbreakers both ways
Constraints marked as dealbreakers (for example about children or smoking) apply to both sides.
Same context
Dating, friends & activities and Desires are separate; both people must have the context switched on.
2 · Then a weighted score of documented factors
Everyone who passed the filters is ranked by these factors. The bars show the default weight of each factor in the total score.
Shared intentionsYou can adjust
30 %How well your long-term intentions and current Now plans line up.
Shared interestsYou can adjust
25 %Interests, sports and values you both listed publicly.
Activity and timingYou can adjust
15 %Compatible Now plans, availability and preferred first dates.
Fairness rules
- Every 5th slot goes to a newer profile, so new members get a fair chance to be seen.
- No more than 3 people in a row from the same area, so results stay diverse.
- Ties are broken by a daily, deterministic shuffle — not by popularity.
- There are no premium or hidden pools: everyone eligible can be shown to everyone eligible.
What is never used
- Popularity, number of likes received or “desirability” scores
- Photos, appearance or attractiveness ratings
- Payments or donations — nothing can buy visibility
- Inferred sensitive traits such as orientation, health or religion
- The contents of your private messages
- Desires data outside the opt-in Desires space
Prefer no personalisation?
You can switch modes at any time in Explore. The hard mutual filters always apply.
Recommended
The weighted score above, with your adjustments.
Nearby
Not personalised: sorted by approximate distance only.
Newest
Not personalised: most recently joined first.
Explanations that stay private
Every recommendation can tell you why it appears, in plain language. For example:
“You both enjoy hiking and want to meet this weekend.”
- Reasons only use information you can already see on the other person's profile.
- They never reveal someone's private preferences, Desires data, exact location or internal safety signals.
- We log which algorithm version produced each recommendation (kept for 30 days) so results can be audited.
What the app learns from you
Occasionally, and always optionally, the app asks short questions such as what caught your attention. Your answers are stored as explicit preferences. Behaviour-based learning is limited to public interests, is opt-in, and never touches sensitive traits.
You can see, edit and delete everything that was learned in You → What have you learned about me. Privacy notice