Three-Step Installation Scheduled in Parallel
A telecom provider sells a three-step installation: Survey, Cable Pull, and Activation. Each work order contains three activities. Survey must finish before Cable Pull starts, with at least 45 minutes between them; Activation must start at least 30 minutes after Cable Pull finishes. All three must be performed by the same technician on the same day. In the current configuration, the activities are independent records with no links. The routing engine schedules them at 08:00, 08:30, and 09:00, often on three different technicians, because it treats each activity as a separate demand. As a result, 12% of installations fail because Activation is attempted before the cable is pulled, and 9% require a second truck roll. The activity types are already defined, but the work order type does not enforce sequence. Dispatchers manually review every order, adding 20 minutes per order. Volume is 600 orders per month and growing.
1. How should the three activities be modeled so the routing engine enforces sequence, lag, and same-technician constraints?
Read the worked answer
Model the three activities as linked activities within the work order and add dependencies rather than relying on dispatcher review. Create a finish-to-start dependency from Survey to Cable Pull with a 45-minute lag, and a finish-to-start dependency from Cable Pull to Activation with a 30-minute lag. Set the same-resource constraint on the dependency chain so all three activities must be assigned to the same technician. Keep the activities under one work order so the grouping is visible and so the dependency chain can be generated from the work order template. Use the activity types already defined, but do not rely on the type name to imply sequence. The routing engine will then schedule Cable Pull no earlier than 45 minutes after Survey completes and Activation no earlier than 30 minutes after Cable Pull completes. If the dependencies are created only after dispatch, the first routing pass will still schedule them in parallel; build the dependencies at order creation, either in the template or in the inbound integration.
2. What should happen when Survey is not completed or is marked not done? How would you prevent Activation from being dispatched?
Read the worked answer
A dependency holds a successor only while the predecessor is incomplete. If Survey is marked not done, the successor must not become eligible just because the predecessor reached a terminal status. Configure a business rule or message scenario that reacts to the not-done reason: set Cable Pull and Activation to cancelled or to a held status with a clear reason, and notify the dispatcher. Do not leave them active, because the routing engine may treat them as ordinary demand. If the business wants to reschedule rather than cancel, move the successors to a pending status and create a new Survey activity; do not let the old Activation remain linked to a failed predecessor. In the mobile workflow, block the completion of Cable Pull until Survey is completed, and block Activation until Cable Pull is completed. Add a daily audit of work orders with a completed successor and an incomplete predecessor; that catches integration gaps.
3. How would you validate the design before full rollout, and which operational metrics would show whether it worked?
Read the worked answer
Run a pilot of 50 to 100 orders in one work zone. Confirm that the dependency chain is created at order creation, that routing assigns all three activities to one technician, and that the gaps are at least 45 and 30 minutes. Test the exception paths: Survey not done, Survey completed late, and a technician cancellation. Check that successors are not dispatched in any of those cases. Compare pilot results with the current baseline: first-time activation success, second truck rolls, manual dispatch review minutes per order, and on-time completion. A successful pilot should reduce failed activations from 12% toward zero and cut the 20 minutes of manual review. After rollout, monitor dependency violations, same-resource violations, second truck rolls, and the percentage of orders where a successor starts before its predecessor completes. Also monitor schedule density: if the lag times make the three activities too long for one day, routing will push the order to the next day, which may be a business trade-off rather than a defect.