Light Rail Extension
Interface management across five packages drove the result. Weekly 30-minute interface forum closed an average of 8 interfaces per week.
- Value
- $410M
- Duration
- 48 months
- Outcome
- Delivered
- Interface register opened week 1
- Drawing release tracked weekly
- Earned schedule beat SPI past 60%
- Interface management is the project on multi-package programmes
- Earned schedule is more honest than SPI past 60%
- Drawing release is a leading indicator of delay
Project context
A 14-kilometre extension to an existing urban light rail network, comprising 11 new stations, two depots, a tunnel section of 1.8 km under an established suburb, an elevated viaduct of 2.6 km across a river floodplain, and a new control and signalling system integrated with the existing network. The contract was an alliance between the state transport authority, a Tier-1 civil contractor, a rail systems integrator, a rolling-stock supplier and an operator. Programme at award was 48 months with a regulated revenue-service start date set by the transport regulator. Total alliance target outturn cost was AUD 620 million (approximately USD 410 million). Five major packages were running in parallel from month 6: civil substructure, viaduct, tunnel, stations, and systems. The single biggest risk on the project from day one was the interface between the five packages — not the work within any one of them.
Standing up the interface register in week one
The alliance leadership team made an unusual decision before mobilisation: the interface register was opened in week one of the project, with named owners on every package, and a baseline set of 86 interface items identified through a three-day cross-package workshop. Most rail programmes open the interface register reactively, six to nine months in, when the first major clash forces it. Opening it in week one meant the team spent the first six months chasing items that did not yet have a real consequence — which felt unproductive — but established the cadence and the discipline before the consequences became expensive. By month 12, when the first major civil-systems interface emerged at the depot, the team had a working process to close it inside three weeks rather than the three months it would have taken in a reactive set-up.
The Tuesday interface forum
Every Tuesday morning from week one to substantial completion, the five package managers and the alliance project director met for 30 minutes. The agenda was fixed: new interface items raised in the past week, items due for closure this week, items overdue, and a single escalation slot for items that needed alliance-level decision. The forum closed an average of eight interface items per week across the life of the project, with a total of 1,840 items raised and 1,820 closed before substantial completion. The 20 remaining items were closed during the testing and commissioning phase. The forum was protected — it ran on Tuesday morning even when the alliance director was travelling, with a deputy chairing. The cultural lesson is that a cadence that runs every week, no matter what, becomes part of the project. A cadence that slips when other things are busy becomes optional, which means it stops working.
Drawing release as a leading indicator
On a programme of this scale, drawing release rates are a more reliable predictor of construction delay than any schedule metric. The team tracked weekly drawing issue counts from the design consultant against a baseline derived from the planned construction sequence. When the drawing release rate dropped to 60% of baseline in month 9, the team identified that two specific design sub-consultants were under-resourced. The remediation was operational: additional resources were brought in within four weeks, and the drawing release rate was back to 95% of baseline by month 12. If the team had waited for the SPI to react, the delay would have been visible in month 14 at the earliest and would have cost approximately six weeks of float to recover. Drawing release leads SPI by roughly 12 to 16 weeks on a project of this kind.
Earned schedule past 60%
SPI as defined in standard EVM becomes increasingly unreliable past 60% complete because the planned-value curve flattens. A project that is genuinely behind can show an SPI close to 1.0 in the final 20% simply because there is so little PV remaining to earn against. Earned schedule, which converts EV back into time units, is a far more honest metric in the late phase. On this project, SPI showed 0.98 at 70% complete while earned schedule showed a 6-week slip. The alliance reported both numbers every month from 50% onwards and the steering committee was trained to read earned schedule as the primary signal. The substantial completion date was achieved with a 9-day variance against a 48-month baseline.
Systems integration and the final 12 months
The final 12 months of a rail programme are dominated by systems integration: signalling, traction power, communications, control systems and rolling-stock interfaces. The team had baked an explicit 14-month systems integration window into the programme, with civil and stations completing 14 months before revenue service. This is longer than most clients accept at award, but on alliance contracts the team has the standing to argue for it. The window was used for static testing, then dynamic testing with empty trains, then trial running with simulated passengers, then a 90-day reliability demonstration before revenue service. Revenue service started on the regulated date with a 98.4% on-time performance in the first month, against a contractual requirement of 95%.
What this project says about multi-package programmes
Multi-package rail programmes succeed or fail on interface management, full stop. The civil works are well understood, the stations are well understood, the systems are well understood. What is consistently under-managed is the boundary between them. The single highest-leverage decision on a programme of this kind is to stand up the interface register and the interface forum in week one, protect them ruthlessly, and use drawing release rate and earned schedule as the leading and lagging indicators. Every other discipline is secondary to that.