The Research
CollabScore™ is built on 398+ verified cross-platform collaboration studies from 2023–2025. Below is exactly how the algorithm works — every factor, every weight, every gate.
398+
verified cases
12
scoring factors
5
failure gates
6
platforms covered
Section 1
Every match is evaluated across 12 independent factors. F1 (Platform Compatibility, 20% weight) is powered directly by the 398+-case research database. The remaining factors use algorithmic logic based on creator economy dynamics.
Final score = Σ(factor_score × weight). Archetype gates apply before output.
Research-backed compat score from our case database for the exact platform pair + niche combination.
Jaccard similarity of content tag sets. Below 30% triggers an archetype gate.
Piecewise penalty curve: 2:1 = –5%, 5:1 = –30%, 10:1 = –75%, >20:1 = –98%.
Live-native × short-form mismatches are heavily penalised. Co-stream between Twitch and Kick scores 1.0.
Base score from hour gap: >12h = 0.10. Same-city bonus enables IRL collab flag.
Historical collaboration outcomes as a predictive signal. Updated post-Phase 3.
Jaccard similarity of language tags. Zero match on a live format triggers Gate 5 penalty.
Cosine similarity of age-bracket distributions. Complementary overlap scores higher than identical.
Match against stated partner-gender preference. Neutral (0.75) when no preference set.
SFW × Adult = hard BLOCKED. MATURE × Adult = 0.50. Any soft block caps the total score.
Engagement rate vs platform benchmark by follower tier. A 2× benchmark = 1.0; 0.5× = 0.45.
8-trigger quadratic score — all 8 conditions needed for meaningful probability. Reported separately.
Section 2
Before the weighted score is returned, 5 hard-stop conditions are checked. These represent structural failure patterns observed in the case studies. Matching any gate caps or blocks your score.
Trigger: F2 < 0.30 (niche overlap below 30%)
Trigger: Ratio > 10:1
Trigger: F6 < 0.20 (e.g. live × short-form)
Trigger: F8 = 0 (SFW × Adult)
Trigger: F5 = 0 AND live collab format
Loading database…
Section 7
Viral outcomes in our dataset shared 8 common conditions. CollabScore™ reports a separate viral probability score computed as P(viral) = (triggers_met / 8)².
The quadratic function means all 8 must be true for meaningful viral probability. 4/8 = 25%. 8/8 = 100%.
Both creators have 1M+ followers
Equal follower tier — within 2:1 ratio
Content format is inherently shareable (comedy, fashion, reaction)
Platform-native format used — not raw repurposing
Trending audio or format hook applied (TikTok)
Simultaneous cross-platform posting on the same day
Pre-promotion at least 72h before the collab
Audience demographics overlap but do not fully duplicate (complementary reach)
Methodology
Data Sources
Public platform metrics, creator-reported outcomes, industry case studies, and media coverage from 2023–2025.
Evidence Quality
Each case assigned a confidence score (0–1). Only cases with confidence ≥ 0.6 are used in aggregate calculations.
Algorithm Weights
Factor weights reflect the assessed relative importance of each signal. Only F1 updates automatically as new research cases are added. Other weights are fixed by design, not statistical fit.
Continuous Updates
Research AI agent runs weekly to add new verified cases. Aggregates rebuild nightly. This page reflects live data.
Apply the research
Every recommendation on Corpfluence runs these 398+ cases and 12 factors against your creator profile in real time.