Scoring methodology
Live platform data + ML scores. Every number on a report maps to one of the sections below.
Live from APIs
- Profile, bio, followers, following
- Likes, comments, shares, saves, views
- Posting frequency (last 30 / 90 days)
- YouTube · X APIs
Modeled and inferred
- RankMint, growth, campaign success - trained ML
- Brand match - embeddings + optional OpenAI upgrade
- Demographics - inferred from location and content niche
- Authenticity - heuristics, not a fraud-vendor API
Model credibility
Dataset
campaign_labels.csv
Rows
403
Accuracy
95.1%
AUC
1.000
F1
0.957
Five scoring engines
Authenticity
Heuristic
- Purchased followers
- Engagement pods
- Bot activity
- Artificial spikes
Growth potential
Modeled
- 90-day follower growth
- Engagement growth
- Audience expansion
Brand match
Embeddings + RAG
- Creator to brand similarity
- pgvector retrieval
- Commerce signal rerank
Campaign success
ML
- Logistic regression
- Engagement and consistency features
- Trained on campaign labels
R
0-100RankMint
Ratefluencer Score
ML composite
- Ranks creators by business impact
- Not follower count alone
- Tier calibration for mega creators
Audience demographics
Age, country, gender on every report
Built from creator location, content niche, and platform benchmarks.
Add brands in Brand workspace · Retrain model with npm run ml:train