Methodology / Matching
How Smart Matching works.
The matching algorithm pairs you with the right person based on skill, goals, timezone, and interaction history. Here is exactly how it makes decisions.
Matching flow
You
Skills, goals, timezone
Matching Engine
6 factors scored
Best match
Ranked by fit score
The problem with random matching
Most peer learning platforms use simple filtering: find someone who teaches what you want to learn, show a list, let users pick. This produces mediocre results because it ignores skill level compatibility, teaching quality, and scheduling realities.
Bad matches waste time for both participants and erode platform trust. A beginner matched with another beginner learns slowly. A learner matched with a teacher in an incompatible timezone never schedules the session. We designed matching to prevent these failures systematically.
Scoring architecture
The matching engine computes a composite score for every potential pairing. Each factor contributes a normalized sub-score. The final score determines ranking in your discovery feed. Matches below a minimum threshold are filtered out entirely.
Scoring runs nightly in batch for the similarity matrix and in real-time when you browse discovery. The nightly batch precomputes expensive calculations like teacher quality indexes. Real-time scoring applies your current filters and availability.
How matches improve over time
Every completed session feeds back into the matching model. If you consistently rate sessions with a particular teaching style highly, the engine learns to weight that pattern. If you cancel sessions with users in distant timezones, the timezone penalty increases for your future matches.
The best matching algorithm is the one that gets better every time you use it.
What we do not do
We do not sell match placement. Paying more does not make you appear higher in discovery. We do not use engagement optimization tricks like showing intentionally imperfect matches to increase browsing time. The engine optimizes for session completion rate and quality, not time spent scrolling.
Matching factors
Skill alignment
The engine finds users whose teaching skills match your learning goals and vice versa. This is the foundational filter. Without skill alignment, no match is proposed regardless of how strong the other signals are.
Skill level compatibility
A teacher should be at least one tier above the learner in the matched skill. Teaching a concept you barely understand produces poor sessions for both participants. The engine enforces a credibility gap between teacher and learner.
Goal complementarity
Users who share aligned goals produce better sessions. If you want to learn Python for data science and a teacher specializes in Python for data science rather than web development, the engine prioritizes that match.
Timezone overlap
Sessions require both participants to be available at the same time. The engine calculates timezone overlap and penalizes matches where scheduling would require either person to join outside reasonable hours.
Session history
Past session quality between two users influences future match scores. Repeat pairings with high ratings are boosted. Pairings with low ratings or cancellations are suppressed. This creates a feedback loop that improves match quality over time.
Language match
Users who share a common language can communicate more effectively. The engine prefers matches with shared language but does not require it, since skill demonstration often transcends language barriers.
Smart Matching FAQ
Yes. The matching engine ranks suggestions, but you can browse the full discovery feed and request a session with anyone. The algorithm surfaces the best matches first, but it does not restrict your choices.
Start proving what you know.
Early access is rolling out for individuals and teams. No credit card, no PDFs — just the things you made, made visible.