Decisions that were slower and dearer before they were right

Replacing three transport systems one customer cohort at a time

When I joined Morrowgate in March 2021, its three divisions moved about 2.3 million consignments a month through three transport management systems, the oldest written in-house in 1998 and hosted under a contract costing A$4.1 million a year. Those systems carried 410 customer connections, from modern message exchanges to spreadsheets emailed each night. Every one of them was somebody's invoice.

Two plans reached the executive committee in mid-2021. Our chief financial officer favoured a single cut-over at Easter 2023, which would retire the hosting contract early and avoid paying for two platforms. I proposed migrating by customer cohort over three to four years, grouping customers by the connections they used and running both platforms until each cohort had invoiced cleanly for two billing cycles. My plan cost about A$11 million more, mostly in extended hosting and parallel running, and it postponed automated rate quoting for small shippers, which the commercial team wanted early. The CFO's objection was fair: a long migration keeps a business distracted. I argued that a failed weekend would cost us customers, not merely money. The board approved A$64 million for the cohort plan in October 2021.

We catalogued every connection before configuring anything, retired 54 duplicates and moved 120 customers to a standard interface. Interstate customers moved first, finishing in November 2023. The port wave in August 2024 went badly: our interface with the stevedores' vehicle booking systems failed at Brisbane for 52 hours, and the team booked 1,900 container movements by hand. Rework took the programme to A$72 million.

Contract logistics completed the migration in May 2025, and the legacy hosting contract ended that June, about two years later than the CFO's plan would have achieved. Customer connections now number 236, and invoice disputes have fallen from 4.1 per cent of invoices to 1.5 per cent. No customer missed an invoice run during the migration, and the next change to any connection is cheaper for both sides.

I would do two things differently. I would test the stevedore interfaces against a full week of real booking volumes rather than a sample day, because port systems belong to other companies and cannot be fixed on our timetable. And I would start cleaning customer master data a year before the programme, not alongside it.

Starting with 24 electric prime movers instead of the 60 we had announced

In February 2023 our board announced a goal of 60 battery-electric prime movers on the Melbourne port shuttle by June 2025, and four months later the local distribution network business quoted A$9.6 million and thirty months to give our depot the 4 MVA supply that fleet would need. The shuttle was the right place to begin: short, predictable trips of about 205 kilometres a day between the wharves, our container park and our western Melbourne rail terminal. The connection was the problem, and the goal was already in our annual report.

Our managing director and our head of sustainability wanted to accept the quote and keep the date, leasing diesel units in the meantime. Our chief operating officer worried that any electric fleet would hurt availability. I recommended a different sequence: 24 trucks on the 1.5 MVA supply we already had, with charging scheduled by software, a 1.6 MWh site battery and 450 kilowatts of rooftop solar, and a revised goal announced at once. What we gave up was the date, and with it part of the emissions reduction we had promised for 2025–26. The board accepted the recommendation in August 2023, and the managing director explained the revision on our results call.

The site works cost A$6.1 million, and all 24 trucks were in service by April 2025. Managed charging has held the depot's peak demand at 1.38 MVA. The trucks average 1.28 kilowatt-hours a kilometre, availability is 93 per cent against 95 for diesel units on the same work, and energy costs 38 per cent less per kilometre than diesel. The fleet avoids about 820,000 litres of diesel a year. One limit proved larger than I had estimated: the batteries add roughly four tonnes of tare mass, so about 18 per cent of shuttle containers are too heavy for the electric combinations.

With eighteen months of measured load data, the network business has offered a 2.5 MVA upgrade for A$5.4 million by late 2027, enough for 20 more trucks. The board will decide in February 2027.

I would share our measured load profile with the network in the first month rather than after a year. I would also design the midday top-up with drivers from the start, because our first schedule clashed with their meal breaks.

Renewing a A$190 million contract on delivery events, not a live feed of our drivers

Early in 2024 the grocery retailer that is our largest contract-logistics customer, worth A$190 million a year to us, asked for a live telematics feed from every vehicle serving its distribution centres as a condition of renewing our five-year contract. The request covered about 310 vehicles, including 85 run by owner-drivers: positions every thirty seconds, speed, harsh-braking alerts and the identity of each driver.

I understood the reason. Its dock schedulers worked from emailed estimates, and our trucks arrived in clusters. But the feed would have sent a continuous record of named people's working days to a company that did not employ them, including owner-drivers whose vehicles we do not own. It would also have blurred responsibility for safety data that we, as the operator, must act on ourselves under the Heavy Vehicle National Law. Our tracking notices to employed drivers, which New South Wales law requires before vehicle tracking begins, said nothing about a customer receiving the data.

I proposed an event feed instead: departure, revised arrival estimates, arrival at gate and dock, temperature excursions for refrigerated loads, and vehicle position only within 40 kilometres of the customer's sites, with no driver identity or speed. Our chief commercial officer feared we would lose the contract on principle, and the division's general manager agreed with him. The customer's supply-chain director rejected the first offer.

We ran a six-week pilot at two distribution centres. Average dock waiting fell from 74 to 41 minutes, because schedulers could plan from estimates that revised themselves. The contract was renewed in July 2024 on the event feed. Agreement cost us a 1.1 per cent rate reduction, about A$2.1 million a year, and A$1.4 million to build the feed. In exchange the customer gained an annual right to audit our fatigue and speed management.

The data schedule we wrote for that renewal has become standard and now covers 52 customer contracts. No driver identity leaves Morrowgate through it, and we have lost at least one smaller tender because of it.

What I would change is the order of conversations. I spoke to the customer before our drivers' representatives, and they learnt of the request second-hand. They supported the outcome, but they should have shaped it.

Removing standing vendor access to cranes and gates, and accepting slower repairs

In mid-2022 our first complete inventory of operational technology at the three intermodal terminals and two container parks found 27 always-on remote connections that equipment and software vendors used to support cranes, reach stackers, gate cameras and terminal software. Several used shared passwords. Two had been installed for projects finished years earlier and were still open.

Vendor access is how equipment gets fixed quickly. A crane supplier's engineer in another state can diagnose a fault in minutes over a live connection, and our terminal managers are measured on train departures. Our general manager of terminals argued that removing standing access would delay repairs and make trains miss their paths on the interstate network, where a missed slot can cost a day. Two vendors said their contracted response times depended on it. They were not wrong.

I recommended replacing every standing connection with brokered sessions: requested by the vendor, approved by the terminal duty manager, limited to named equipment, recorded and closed automatically after four hours. We separated terminal equipment networks from the corporate network at the same time. The programme cost A$4.3 million, and we renegotiated support terms with nine vendors, two of whom charged more for the change.

The cost appeared quickly. In the first six months average restoration time for equipment faults rose from 3.2 hours to 4.7 hours, and in October 2022 two trains left our Sydney terminal late after a gantry crane fault took longer to diagnose. The general manager reported both to the executive committee, rightly. We stocked critical spares on site and trained eleven of our own technicians in first-line diagnosis, and by December 2023 restoration time was back to 3.5 hours.

In November 2023 a cyber incident stopped landside operations at one of the large stevedores for several days, and our board asked how exposed our own terminals were. I could show every remote session into our equipment for the previous year, who approved it and what it touched. That answer existed because we had accepted slower repairs.

I would change one thing. I removed access in the middle of support contracts, which gave vendors grounds to reprice. I would now write brokered access into each contract at renewal, and fund the spares and training before switching anything off.