AI analyzed 18.4 million predictions today across 247 production models. Sentiment Analysis v1.4 has the least-explainable predictions this week — its average SHAP confidence dropped 11 points against baseline, an explanation-quality anomaly worth reviewing. Churn Prediction v3.2 remains the most explainable model in the fleet at a 94.2 quality score. Recommendation: schedule an explainability audit for Sentiment Analysis v1.4 before its next retraining cycle.

Select a model to inspect

Inspecting Churn Prediction v3.2 — most recent scored prediction, account #48213.

This prediction was primarily driven by 3 missed payments in the last 60 days, which increased churn risk by 34%. A declining usage trend over the past 30 days contributed an additional 18% risk, partially offset by a strong Net Promoter Score, which reduced predicted churn risk by 9%. Overall, the model is 91% confident in this churn classification.

Explainability Reports