Think Transit was engaged by GPM to analyse the current signal maintenance scheduling data and identify opportunities for optimisation. Using the Signalling TMP schedule and 12 months of work order data covering the full metropolitan network (16,575 work orders across 44 work groups and 14 depots), we built a series of analytical models and interactive visualisation tools to examine how work is currently assigned.
1. The scheduling system applies no geographic optimisation
The data is consistent with system-generated scheduling from the ERP, with planned dates placed at the midpoint of each maintenance window and work groups assigned from the asset register. There is no evidence of geographic optimisation — work at the same location is spread across multiple days despite overlapping maintenance windows.
2. 73% of location visits could be eliminated
The same team visits the same location on multiple separate days when the tolerance windows overlap and all work could be batched into fewer visits. Across the network, 4,226 annual location visits could be reduced to approximately 1,121 — a reduction of 3,105 visits.
3. 31% of visits are for a single work order
1,321 times per year, a team travels to a location to perform just one job. 96% of these (1,270) could be combined with another visit to the same location by the same team.
4. Multiple teams attend the same location on the same day
252 times per year (excluding specialist groups), two or more regular signals sections attend the same station on the same day performing similar or identical work. In the clearest cases, three teams from two different depots perform the same enclosure inspection at the same station.
5. Work groups are highly interchangeable
53% of all work orders are for standard jobs performed by 25 or more of the 36 regular signals sections. There are few technical barriers to consolidating work within or across depot sections.
| Case Study | Location | Finding |
|---|---|---|
| Aspendale | Frankston line | 4 visits in 5 days for level crossing inspections on the same rail line, walking distance apart. 41 km of unnecessary travel in one week. |
| Southern Cross | CBD | 13 visits in 14 days — flexible relay inspections drip-fed daily instead of batched. 57 unnecessary mobilisations per year at Melbourne's busiest station. |
| Burnley | Inner metro | 3 teams from 2 depots all doing the same enclosure inspection on the same day. Repeats 9 times per year. |
| Caulfield | Inner metro | Same-depot sections both at Caulfield — one team's entire day is 3 WOs the other team could absorb. Repeats 14 times per year. |
The current scheduling approach treats each work order independently. It does not consider: - What other work exists at the same location within the tolerance window - What other teams are scheduled at the same location on the same day - The geographic sequence of a team's daily assignments
An intelligent scheduling system that factors in these three dimensions would significantly reduce unnecessary travel, eliminate redundant mobilisations, and free team capacity for productive work.
1. Tolerance window batching Implement location-based grouping when generating the maintenance schedule. Before placing a work order at its midpoint, check whether the same team has other work at that location within the tolerance window and batch them onto the same day.
This is the highest-value change and addresses the largest source of inefficiency (73% of visits). It requires no organisational change — only a modification to how planned dates are calculated.
2. Cross-section visibility at overlap locations For stations that sit on depot boundaries (Burnley, South Yarra, Footscray, Clifton Hill, Southern Cross), implement a check that flags when multiple sections are scheduled at the same location on the same day. A simple daily report would allow schedulers to manually consolidate.
3. Single-WO visit review Generate a weekly report of planned single-work-order visits and flag those where another visit to the same location exists within the tolerance window. These are the easiest consolidation decisions — they require no judgment about capacity, just moving one job to an existing visit.
4. Geographic scheduling engine Develop or procure a scheduling tool that considers asset location, team base depot, tolerance windows, and team capability when generating the maintenance plan. The tool should: - Group work by geographic proximity within tolerance windows - Assign work to teams based on depot distance and capability - Optimise the daily sequence of locations using nearest-neighbour or similar routing - Flag cross-depot overlaps at boundary stations
5. Capacity modelling The current analysis does not model team capacity (how many work orders a team can complete in a day). Adding duration estimates per standard job would allow the scheduling engine to balance workload across days while maximising geographic batching.
6. Dynamic rescheduling Move from a static annual schedule to a rolling schedule that adjusts as work is completed, faults occur, or conditions change. Tolerance windows provide natural flexibility — a dynamic system could continuously re-optimise the forward schedule as each day's work is completed.
7. Cross-depot optimisation The current depot territory model assigns assets to a fixed depot. For boundary locations, a cross-depot optimisation could assign work to whichever depot already has a team nearby on that day, regardless of the traditional territory boundary.
| Phase | Timeframe | Focus | Estimated Effort |
|---|---|---|---|
| 1. Immediate | 0-3 months | Tolerance window batching, single-WO visit reports, overlap flagging | Configuration/reporting changes |
| 2. Build | 3-12 months | Geographic scheduling engine, capacity modelling, cross-section coordination | Software development or procurement |
| 3. Optimise | 12+ months | Dynamic rescheduling, cross-depot optimisation, continuous improvement | Ongoing refinement |
Phase 1 can be implemented with minimal investment using the existing scheduling system's configuration and standard reporting. It addresses the largest single source of inefficiency (tolerance window batching) and provides immediate, measurable results.
Phase 2 requires either development of a bespoke scheduling tool or procurement and configuration of an existing workforce scheduling platform. The interactive tools developed as part of this POC demonstrate the analytical framework that such a system would need.
Phase 3 represents the transition from periodic schedule generation to continuous optimisation, which requires integration with the work management system and real-time feedback on work completion.
Think Transit is available to support any of these next steps, from business case development through to implementation.
The following enhancements could be delivered to deepen the analysis and strengthen the business case:
| Enhancement | Description | Value |
|---|---|---|
| Real road routing | Replace straight-line distance estimates with actual road routes and distances using OSRM, displayed as vector paths on the daily map | Accurate travel cost quantification; visually compelling route visualisation |
| Historic duration analysis | Integrate the 2017-2026 historic completion data to determine actual labour hours per standard job | Enables realistic daily capacity modelling and validated workload limits for the optimised schedule |
| Optimised schedule generation | Generate a complete optimised schedule (V2) with tolerance window batching and capacity constraints, presented as a side-by-side comparison with the current schedule | Demonstrates the tangible before/after improvement; quantifies exact trips and km saved |
| Cost model | Build a cost model incorporating labour rates, vehicle costs, access/protection arrangements, and admin overhead per mobilisation | Converts trip reductions into a dollar value for the business case |
| Depot-level pilot analysis | Deep-dive into one depot (e.g. Carrum or Dandenong) with a fully validated optimised schedule and detailed savings breakdown | Provides a concrete, auditable pilot proposal for Phase 1 implementation |
| Cross-depot boundary analysis | Model alternative territory boundaries and work assignments for stations serviced by multiple depots | Quantifies the additional savings from cross-depot coordination |