Six appointments between the site docket and the board paper

Mandate

Make the platform dependable during the site morning, meet customers' security and data-residency obligations, and keep cloud costs proportionate as the company grows.

Outcomes

• Enforced separation of customer data in the database after a near miss in 2023
• Reduced cloud spending from 11.6 to 8.3 per cent of revenue between the 2024 and 2026 financial years
• Built an Australian standby region with weekday-morning availability of 99.96 per cent in 2026
• Introduced a published technical-debt register reviewed by the board each half-year

Decisions in context

I became chief technology officer when Colin Varnstead, our co-founder and first technical lead, left to start a separate venture. The company then had A$26.1 million of revenue and forty-one engineers. Revenue has since grown to A$47.2 million while the engineering organisation has grown more slowly, to fifty-eight. That ratio is deliberate: I would rather fund fewer teams properly than spread maintenance thinly. The role changed after the large Australian breaches of late 2022, when security moved from a technical subject to a standing board item. I now report quarterly to the audit and risk committee on reliability, security maturity, residency obligations and cloud spending. All three decisions described on my Impact page were taken in these five years. Each required me to recommend something that cost money or customers, and to explain the trade-off to people who were not engineers.

Reason for the move

Current appointment; seeking a first non-executive role alongside it.

Mandate

Turn a founder-led engineering group into teams that could release safely without depending on a handful of people.

Outcomes

• Moved releases from fortnightly evenings to daily daytime deployments with automated checks
• Introduced on-call rotations and written incident reviews for every customer-facing outage
• Supported the move from one state's safety paperwork to five states' permit formats

Decisions in context

Quorrindale's growth in these years came mainly from builders working across state borders, which forced the product to handle different safety paperwork and permit conventions. I organised engineering into five teams around parts of the site day: the gate, the paperwork, the diary, defects and the platform beneath them. We moved from evening releases, which one senior engineer supervised, to daily daytime deployments with automated checks, so that a mistake was caught by colleagues who were awake. I also introduced written incident reviews that described the timeline before discussing any cause. The period had a real failure as well. In 2018 I listed weak separation between customer accounts as a risk and twice allowed it to slip behind features. That decision came back to us in 2023, and it shapes how I now handle deferred risks with the board.

Reason for the move

Appointed chief technology officer when the co-founder left the company.

Mandate

Stabilise a product that builders liked but that failed under growing morning load.

Outcomes

• Introduced tested, off-site backups and a written recovery procedure
• Added monitoring that started before the first sites opened each morning
• Reduced weekday-morning incidents from eleven in 2014 to two in 2016

Decisions in context

I joined a company with an affectionate customer base and a fragile system. The sign-in product had been built quickly by the founders and a few contractors, and it slowed badly whenever several large sites opened at once. My first year was spent on unglamorous foundations: backups that had actually been restored, a release process that did not depend on one person, and monitoring that started at 5.30 am rather than when the office opened. I also visited around thirty sites to watch the morning sign-in in person. Those visits changed our priorities more than any dashboard. I learnt to recruit engineers who were curious about construction, because the best fixes came from people who understood why a foreman would give up on a screen after ninety seconds. Weekday-morning incidents fell from eleven in 2014 to two in 2016.

Reason for the move

Promoted to director of engineering as the team and customer base grew.

Mandate

Keep hosted customer systems available and make recovery repeatable rather than heroic.

Outcomes

• Wrote the company's first standard format for incident reviews
• Ran quarterly recovery tests for hospital and council customers
• Reduced repeat incidents by tracking every review action to completion

Decisions in context

I joined Wollerby to learn what happened to software after developers handed it over. For five years I was on the on-call rota, first as an engineer and from 2012 leading a team of eight. Our customers included retailers, councils and two hospitals, each with a different idea of what an acceptable outage meant. I learnt to separate the timeline of an incident from the argument about fault, and to write the timeline first. I also learnt that a recovery procedure nobody has rehearsed is a hope rather than a plan. We began quarterly recovery tests for the hospital and council customers, which were uncomfortable the first few times and steadily less so. By 2014 I wanted to work on a product of my own again, and a role at a construction-software company returned me to the site sheds where I had started.

Reason for the move

Wanted to build a product rather than host other people's, and to return to construction.

Mandate

Build and maintain inspection scheduling and building-permit records with a defensible audit history.

Outcomes

• Redesigned the inspection record so that every change kept its author and reason
• Led the inspections product through adoption by eighteen councils
• Cut the backlog of customer-reported defects from 140 to 35 items

Decisions in context

Council software taught me that a record's history is part of the product. A building permit or an inspection result might be challenged years later, and a council officer needed to show who had changed what and why. I redesigned our inspection record so that every alteration kept its author, time and reason, which made the product slower to build but much easier to defend. From 2006 I led six developers on the inspections product as eighteen councils adopted it. I also learnt how a public-sector customer buys and uses software: carefully, with long procurement cycles and little tolerance for surprises. Those habits still shape the way I discuss changes with government agencies that require our builders to use Quorrindale. I left because I realised I understood very little about running software in production, and I wanted to learn it properly.

Reason for the move

Moved to a hosting company to learn operations and recovery at first hand.

Mandate

Replace paper plant-hire and timesheet records with systems that site supervisors would actually use.

Outcomes

• Built a plant-hire system that allocated machine hours to jobs and clients
• Replaced weekly paper timesheets for about 200 site workers
• Reduced disputed plant charges with clients by recording hours at the site office

Decisions in context

My first employer was a family-owned civil contractor building roads, drainage and subdivisions around Newcastle. I was the second of two in-house developers, reporting to the financial controller. Our main task was replacing paper plant-hire dockets and weekly timesheets. The supervisors who filled them in were tired at the end of a long day, and any screen slower than the paper simply went unused. I spent many afternoons in site sheds watching how people actually recorded hours, and I rewrote the entry screens three times. By 2003 the plant-hire system allocated machine hours to jobs and clients, and disputed charges had fallen noticeably. The job gave me the question I still ask before most decisions: what does the person on site need in the next few minutes? I left to work in a software company, where I could learn from more experienced engineers.

Reason for the move

Moved to Melbourne to work for a software company rather than an in-house team.