π’ 1️⃣ MOST USED METRICS IN INDUSTRY
Let’s be honest.
Out of 41 metrics, only ~12 are heavily used daily in real companies.
π΄ Classification (Industry Core)
-
Accuracy (only when balanced)
-
Precision
-
Recall
-
F1 Score
-
ROC-AUC
-
PR-AUC (very important in imbalance)
-
Log Loss
π΅ Regression (Industry Core)
-
MAE
-
RMSE
-
R²
-
MAPE (forecasting)
π’ Production Monitoring
-
Drift metrics (PSI / KS)
-
Calibration (Brier Score)
-
Latency
π‘ Computer Vision
-
mAP
-
IoU
π£ NLP
-
BLEU / ROUGE (less in production)
-
Perplexity (LLMs)
π 80% of real-world ML uses:
Precision, Recall, F1, ROC-AUC, MAE, RMSE, R²