Imran Anwar

Home / Notes

Workday Message Queue Service vs scheduled integrations

Use a scheduled integration when near-real-time delivery is not needed and the downstream system is reliably available. Use Message Queue Service when events cannot wait for the next batch window or the receiver is sometimes down. MQS removes the lost-event problem but delivers at-least-once, so the receiving side must handle duplicates.

When polling is fine

Scheduled polling means an EIB or RaaS report running on an interval. It is simple to build, simple to monitor and easy to explain to a support team. If the business is happy with data that is an hour or a day old, and the target system is up when the job runs, polling is the right answer. Do not add a queue to a flow that does not need one.

Where it breaks

Polling struggles in two cases. The first is latency: an event happens, and nothing moves until the next run. The second is availability. If the downstream system has intermittent availability, a run can fail or be skipped, and the event waits for the next window. You end up writing catch-up logic, date windows and manual reruns. That logic is where the bugs live.

What MQS changes

Workday Message Queue Service enqueues each event and retries delivery until it is acknowledged. It does not matter whether the downstream system was up when the event occurred. Paired with Studio Listener Service for event-driven inbound triggers, it replaces clock-driven integration with event-driven delivery. I built this pattern at Cognizant, with at-least-once delivery and idempotent consumers.

The idempotency cost

Delivery is at-least-once, never exactly-once. A queue that retries until it gets an acknowledgement will sometimes deliver a message twice, for example when the consumer processed it but the acknowledgement was lost. Retrying is what prevents lost messages, so duplicates are the price.

The fix belongs in the consumer. Make processing idempotent, so handling a message a second time gives the same result as handling it once. The usual approaches:

  • Key each operation on a stable event identifier.
  • Use upserts instead of blind inserts.
  • Record which identifiers have already been applied.

This is real design work, and it moves the complexity from the Workday side to the receiving side. Teams that skip it find out the first time a retry creates a duplicate record. Budget for it before you choose MQS.

A short decision guide

  1. Can the business accept data that is as old as the polling interval? Stay with a schedule.
  2. Is the downstream system reliably available when the job runs? If yes, a schedule still works.
  3. Do events need to arrive soon, even after an outage? Use MQS.
  4. Can the consumer be made idempotent? If not, fix that first, or do not use MQS.

If you want to see the retry and outage behaviour before you decide, try the MQS Simulator.

Imran Anwar, Principal Workday Integration Consultant. Published 2026-11-16.