Tolerance Window Analysis

Key Finding

By using the existing tolerance windows (min planned to max planned) to consolidate visits, 73% of location visits could be eliminated — reducing from 4,226 visits down to 1,121 across the year.

This means the same team is currently visiting the same location on average 3-4 separate days when, within the allowable maintenance windows, all that work could be batched into a single visit.

Current vs Optimised Visits

Metric Current Optimised Saving
Total location visits (group+location+day) 4,226 1,121 3,105 (73%)
Unique group+location combos with savings 268

Savings by Depot

Depot Current visits Optimised Saved % reduction
Flinders Street 571 121 450 79%
Clifton Hill 456 129 327 72%
Newport 424 98 326 77%
Carrum 395 119 276 70%
East Malvern 377 101 276 73%
South Kensington 377 108 269 71%
Sunshine 335 71 264 79%
Ringwood 330 81 249 75%
Dandenong 316 77 239 76%
Caulfield 274 89 185 68%
North Melbourne 192 39 153 80%
Spencer Street 101 17 84 83%
MURL 14 7 7 50%

Worst Examples

The most extreme cases of repeated visits to the same location by the same team:

Work Group Location Current visits Could be Saved WOs
Flinders Street 1203 Flinders St A 89 11 78 418
Spencer Street 1101 Southern Cross 87 9 78 316
Flinders Street 1201 Flinders St E 85 8 77 287
Newport 2201 Newport 80 10 70 300
North Melbourne 1302 North Melbourne 73 10 63 291
Dandenong 4301 Westall 71 9 62 333
Clifton Hill 3301 Epping 68 10 58 254
Ringwood 3201 Box Hill 35 5 30 115

For example, Flinders Street Section 1203 visits Flinders St A 89 separate times across the year to do 418 work orders. By batching work within the tolerance windows, this could be reduced to just 11 visits.

How This Works

Each work order has a tolerance window defined by min_planned and max_planned. The current scheduling system places the planned date at the exact midpoint of this window, with no consideration of other work at the same location.

The analysis uses a greedy merge approach: starting with the earliest visit, it checks whether the next visit's tolerance window overlaps. If it does, both visits could be done on a single day. This continues until no more merges are possible, at which point a new visit is required.

Important Caveats

  1. Capacity constraints not modelled — batching 89 visits into 11 means some days would have very high work order counts. A team may not be able to complete 40+ WOs in a single day. Practical limits would need to be applied.

  2. We assume no optimisation has been applied — the planned dates sit at the exact midpoint of the tolerance window (confirmed for 96% of WOs), suggesting the scheduling system applies no geographic or workload optimisation.

  3. Travel time is not included — this analysis only looks at visit consolidation within a single group+location. The additional benefit of reduced travel between locations is not quantified here but would compound these savings.

  4. Different job types may have different durations — batching a trainstop inspection with a point mechanism overhaul onto the same day only works if there's enough time. Job durations are not available in the current dataset.

Implication

This finding represents the second layer of optimisation opportunity beyond the overlap analysis. Even without changing which team does the work (the overlap problem), simply rescheduling work within the existing tolerance windows to batch visits by location could eliminate approximately 73% of location visits, with direct reductions in travel time, access/protection requirements, and administrative overhead.

Combined with the overlap consolidation (reducing multi-team visits to the same location), the total efficiency gain from an optimised scheduling approach would be substantial.