Three decisions and what each one cost us

Running the old receival system through one more harvest

Our receival and grower-accounting system was written in 2003 by a small software house whose founder retired in 2019. It ran the weighbridges, recorded sample results, held each grower's stock by grade and produced settlement statements. In 2022 the board approved a A$14.6 million programme to replace it with a hosted receival platform, with cutover planned for September 2023, a month before harvest.

By July 2023 testing showed problems. Sampling-instrument integration was incomplete at nine of fourteen sites, and about six per cent of grower records had conflicting entity or bank details that needed checking with the growers themselves. The general manager of operations wanted to proceed, partly because new handheld sampling instruments had already been bought, and the vendor was confident. The old system's servers were also out of support. Our chief financial officer was concerned about the cost of any delay.

I recommended deferring cutover to March 2024, after harvest, when receivals are light. The trade-off was explicit in the board paper: A$2.6 million for extended contractor support, two replacement servers with spares, and vendor standby fees, plus a season in which the new instruments sat unused. I argued that a failed cutover during harvest would hold trucks at every site at once, with fatigue and road-safety consequences that no budget saving justified. The chief executive supported the recommendation; operations accepted it but did not agree with it.

The 2023–24 harvest received 2.8 million tonnes. One disk failure at a site stopped receivals for 52 minutes, and the paper ticket process carried the queue. The March 2024 cutover took a weekend; reconciliation took three more days, and 140 of about 4,300 grower statements needed correction within a week.

The harvest change freeze and site-by-site readiness testing with live instruments are now standard. I would start cleansing grower records a year earlier than we did, and I would have checked server support dates before the programme was approved rather than discovering them during testing.

Introducing receival booking slots and accepting lost tonnes

At our six busiest sites, average peak queuing during the 2023–24 harvest was 2 hours 50 minutes. On eleven days, queues at two sites extended onto the highway shoulder. Queue time generally counts as work time under the fatigue rules, and as a receiver we are part of the chain of responsibility. The new receival platform could support booking, so in May 2024 I proposed time slots for the 2024–25 harvest.

The objections were serious. Growers' representatives valued the open gate: harvest depends on weather, and a booking can be wrong by the time the header stops. Our trading subsidiary feared that growers would deliver to competing sites without a booking system. The proposal cost A$2.3 million to build and about A$400,000 a year to run.

We chose a compromise rather than compulsory slots. Booked trucks used a dedicated lane; unbooked trucks were still received in a separate queue. Bookings could be made by app, text message or telephone. The safety manager and the general manager of operations co-sponsored the change, and the chief executive accepted that receivals might fall at those sites.

They did. Average peak queuing at the six sites fell to 1 hour 5 minutes, and highway-shoulder queuing occurred on one day, after a power failure. Receivals at those sites finished 4.8 per cent below plan. Some tonnage moved to our other sites, but the trading subsidiary estimated that about 110,000 tonnes went to competitors, with roughly A$1.1 million of margin forgone. I report that figure alongside the queue data each season.

Queue time is now a published site measure reviewed weekly by the safety committee during harvest. I would involve subcontracted carriers earlier next time. Our design assumed grower-owned trucks making single trips; carriers running several farms a day could not hold a fixed slot. In the third week we added a ninety-minute carrier window, which worked, but the first fortnight cost those carriers patience that we had to rebuild.

Closing our trading arm's early view of the shipping stem

In 2021 I commissioned an access review across our main systems. It found that staff at Ninderbrook Grain Marketing, our trading subsidiary, had read access to the terminal's nomination system, including other exporters' pending vessel nominations before they were published to the shipping stem. The access came from a shared role created when the system was built in 2012. We found no evidence that it had been used to gain an advantage, but the possibility mattered: the industry code governing bulk wheat port terminals expects an operator not to favour its own trading business.

The decision was how far to go. The trading subsidiary's general manager argued for a narrow fix, removing a few screens, because his team used the same system to plan their own vessels and any separation would slow them. Our legal advisers set out the obligations; the conclusions were theirs and the board's. I recommended a fuller rebuild: separate roles, a nomination workflow in which every exporter's information became visible to all at the moment of publication, logging of every access, and quarterly review by someone outside both the terminal and the trading teams.

The work cost A$780,000 and took seven months. Our traders now plan from the published stem like any other exporter, which delays their internal vessel planning by about a day. Some trading staff felt the change implied wrongdoing. The chief executive and I met them to explain that it reflected a design flaw, not a finding about anyone's conduct.

Since March 2022 the access logs show no cross-access, and an external assurance review in 2023 confirmed the separation. Third-party exporters' share of terminal throughput rose from 38 per cent in the 2021 financial year to 44 per cent in 2025; crop size and exporters' own programmes moved too, so I do not attribute that change to the controls. I would have included the terminal systems in my first access review in 2020. I prioritised finance systems, and that sequencing left the gap open for another year.