EvalGate - Model Quality Dashboard

Published automatically by the CD pipeline after every gated deployment. Latest model: 6044927 trained 2026-10-05 09:57:13 UTC. Model card.

0.8386Test ROC-AUC (95% CI 0.816 to 0.859)

0.025Calibration error (ECE)

+0.0000AUC vs deployed model (95% CI +0.000 to +0.000)

42Release checks enforced, all passed

11Gated deployments recorded

Release gate (enforced in CI - any FAIL blocks deployment)

Group	Check	Value	Required	Gate
quality	roc_auc_test	0.8386	>= 0.82	PASS
quality	accuracy_test	0.7960	>= 0.75	PASS
quality	roc_auc_cv_mean	0.8472	>= 0.82	PASS
quality	cv_test_gap	0.0086	<= 0.03	PASS
calibration	ece	0.0254	<= 0.05	PASS
calibration	brier	0.1390	<= 0.16	PASS
slices	gender=Female	0.8306 (n=681)	>= 0.65	PASS
slices	gender=Male	0.8446 (n=726)	>= 0.65	PASS
slices	SeniorCitizen=0	0.8423 (n=1175)	>= 0.65	PASS
slices	SeniorCitizen=1	0.7829 (n=232)	>= 0.65	PASS
slices	Contract=Month-to-month	0.7574 (n=790)	>= 0.65	PASS
slices	Contract=One year	0.7013 (n=290)	>= 0.65	PASS
slices	Contract=Two year	0.8276 (n=327)	>= 0.65	PASS
slices	tenure_band=0-12m	0.7667 (n=451)	>= 0.65	PASS
slices	tenure_band=13-24m	0.7923 (n=206)	>= 0.65	PASS
slices	tenure_band=25-48m	0.7773 (n=315)	>= 0.65	PASS
slices	tenure_band=49m+	0.8320 (n=435)	>= 0.65	PASS
slices	InternetService=DSL	0.7905 (n=482)	>= 0.65	PASS
slices	InternetService=Fiber optic	0.7731 (n=613)	>= 0.65	PASS
slices	InternetService=No	0.8900 (n=312)	>= 0.65	PASS
fairness	gender AUC gap	0.0140	<= 0.08	PASS
fairness	SeniorCitizen AUC gap	0.0594	<= 0.08	PASS
champion	non-inferior to deployed model	+0.0000 [+0.0000, +0.0000]	lower 95% bound >= -0.01	PASS
drift (held-out vs train)	PSI MonthlyCharges	0.0108	<= 0.25	PASS
drift (held-out vs train)	PSI tenure	0.0091	<= 0.25	PASS
drift (held-out vs train)	PSI TotalCharges	0.0050	<= 0.25	PASS
drift (held-out vs train)	PSI OnlineBackup	0.0035	<= 0.25	PASS
drift (held-out vs train)	PSI TechSupport	0.0027	<= 0.25	PASS
drift (held-out vs train)	PSI StreamingTV	0.0022	<= 0.25	PASS
drift (held-out vs train)	PSI Partner	0.0013	<= 0.25	PASS
drift (held-out vs train)	PSI PaymentMethod	0.0013	<= 0.25	PASS
drift (held-out vs train)	PSI gender	0.0008	<= 0.25	PASS
drift (held-out vs train)	PSI Contract	0.0007	<= 0.25	PASS
drift (held-out vs train)	PSI PaperlessBilling	0.0007	<= 0.25	PASS
drift (held-out vs train)	PSI MultipleLines	0.0005	<= 0.25	PASS
drift (held-out vs train)	PSI DeviceProtection	0.0005	<= 0.25	PASS
drift (held-out vs train)	PSI StreamingMovies	0.0005	<= 0.25	PASS
drift (held-out vs train)	PSI InternetService	0.0003	<= 0.25	PASS
drift (held-out vs train)	PSI OnlineSecurity	0.0003	<= 0.25	PASS
drift (held-out vs train)	PSI SeniorCitizen	0.0001	<= 0.25	PASS
drift (held-out vs train)	PSI Dependents	0.0001	<= 0.25	PASS
drift (held-out vs train)	PSI PhoneService	0.0000	<= 0.25	PASS

Calibration

Each dot is a bin of customers: the model's average predicted churn probability against the share who actually churned. On the dashed diagonal, a predicted 30% means 30% churn. ECE 0.0254, Brier 0.1390.

Test ROC-AUC across deployments

Deployment history

Trained (UTC)	Commit	Test AUC	CV AUC	Accuracy	ECE
2026-10-05 09:57:13	`6044927`	0.8386	0.8472 +/- 0.0033	0.7960	0.0254
2026-09-28 09:18:15	`6044927`	0.8386	0.8472 +/- 0.0033	0.7960	0.0254
2026-09-24 19:58:40	`6044927`	0.8386	0.8472 +/- 0.0033	0.7960	0.0254
2026-09-24 17:50:05	`c4e58df`	0.8386	0.8472 +/- 0.0033	0.7960	0.0254
2026-09-21 08:31:38	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-09-14 08:29:55	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-09-07 07:50:47	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-08-31 09:15:54	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-08-24 03:51:40	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-08-17 03:46:03	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-
2026-08-13 16:52:18	`016e7a7`	0.8386	0.8472 +/- 0.0033	0.7960	-

Pipeline: push -> tests -> train -> release gate -> artifact -> deploy. Policy in gate.toml. Source: github.com/PSPMANI/evalgate | By Pathi Manikanta