A transit authority must score next-day ridership for all routes overnight without interactive latency requirements. Which endpoint pattern fits?

Running a nightly batch managed endpoint job to score ridership for all routes is correct when the transit authority needs next-day scores without interactive latency requirements. Batch endpoints (or batch deployment jobs) process large scoring sets on a schedule, optimize for throughput and cost, and fit overnight planning cycles. Forcing online request patterns onto an all-routes bulk workload adds needless complexity and cost.

Running a nightly managed online endpoint deployment to score all routes fails because a persistent online deployment built for real-time requests is not the throughput-optimized, scheduled path the all-routes overnight volume needs. Deploying a Kubernetes online endpoint deployment to score all routes nightly fails because it adds cluster management for low-latency serving when the scenario has no interactive latency requirement for overnight bulk scoring. Creating a separate batch endpoint job for each route to score nightly fails because splitting one all-routes workload into many per-route jobs adds unnecessary scheduling and monitoring overhead when a single batch job can score all routes.

Exam caveat: batch jobs still need data readiness checks, failure alerts, and output landings that downstream planners consume before morning operations. Do not confuse batch with “no monitoring.” Operational check: configure a nightly batch managed scoring job over all routes, validate output completeness and latency-to-data-availability, alert on partial failures, and confirm planners receive scores before service adjustments.