The work behind the numbers.

As Academic Program Coordinator at IBT Education, I owned student-outcome analysis across 50+ IB subjects. I was also tutoring four of them myself at the time, including Spanish, where I was the company's first tutor.

The signal came from Spanish, a subject I know deeply, so weak results there didn't sit right. I pulled the outcome data across cohorts and found the same shape in the language subjects: the problem tracked with tutor-to-student ratios, not student ability. Spanish and French were carrying around 8.5 students per tutor against roughly 4 elsewhere (Fig. 1.1). I wrote it up for management with a specific fix: reallocate two tutors to the stretched language subjects, refresh their materials, and add subject-specific tutor guides.

Management implemented the changes. For that same cohort, top-band (band 7) attainment across Spanish and French rose from 27% at their mid-year trial exams to 59% at their November finals, and the share finishing at band 5 or below fell from about a third to roughly one in ten. The one-off analysis became a recurring, program-wide data-review process.

It's how I work everywhere: notice what doesn't fit, ask why before deciding what to do, and let the data point to the real lever, then hand back something a team can act on.

At ANZ I resolved small-business client cases in a regulated banking environment, where every resolution has to be correct, compliant, and defensible on audit. Four months in, KPI targets rose from a 30-case daily benchmark toward 45+. The volume wasn't the real problem - the real cost was the time each case took, because the policy and process knowledge needed to resolve them lived in people's heads and across scattered sources.

I treated it as an analysis problem before a speed problem. I mapped where handling time actually went, then built a working reference set of policy and process notes (Fig. 2.1), a single, structured source that standardised how complex scenarios got resolved and sharpened which cases to prioritise. I kept the documentation audit-ready and engaged cross-functional stakeholders to align on compliant resolutions, so the reference set held up under governance, not just at my own desk.

Per-case handling time fell and throughput rose to 45+ cases a day, peaking at 60, roughly a 50% increase on the original benchmark, while I held one of the lowest case-escalation rates on the team and sustained accuracy as volume climbed. The reference set became shared infrastructure: structured documentation that supported knowledge transfer across the team, not a personal shortcut.

This is the PMO half of how I work. Finding the real lever is step one; the value is in turning it into a repeatable system with the governance, documentation, and stakeholder alignment that lets other people rely on it. Process improvement that survives an audit and outlives the person who built it.

A 9,866-word group investigation for BUSM4415 (Industry Project Investigation) into stakeholder engagement on the $1.6B Caulfield to Dandenong Level Crossing Removal Project, focused on construction disruption for traders in Noble Park's Douglas & Ian Street precinct between 2016 and 2018.

I built a desk-based case study around one question: did engagement actually shape how disruption was managed, or mainly justify decisions already made? I tested it against three independent sources: the Victorian Auditor-General's audit, a pre-construction survey of 144 traders, and the minutes of the local community liaison groups.

Found four linked weaknesses: high engagement activity but only 17% pre-construction awareness among rail users, engagement front-loaded before construction and weakest during the disruption itself (Fig. 3.2), no measurement of commercial impact on traders, and the people accountable for community outcomes had no say over construction decisions. The clearest evidence of what was left unresolved: a $3.5M trader revitalisation package committed three years after completion. We delivered five recommendations, including an outcome-oriented evaluation framework and a unified cross-agency Community Outcomes Director role.

Engagement is easy to measure by activity count and easy to get wrong that way. I know how to separate genuine influence on decisions from documentation of decisions already made, and to turn that gap into recommendations a delivery authority could actually implement.

Two linked reports to the board of a company weighing a 20-year deal to build a water-treatment plant, run it, then hand it over. It would cost $85M to build, funded 75% by loans at 12% and 25% by shareholders, with income locked in through three long-term supply contracts.

I built the full 20-year financial model in Excel, then wrote a second report whose job was to pull it apart. That second pass caught that my first report had judged the project against the wrong rate of return. I corrected it, re-checked year by year whether the plant could cover its loan repayments, ran 10,000 simulated scenarios on revenue and costs, and paired each major risk with a specific safeguard.

The project's 13.68% return lands between the two rates it could fairly be judged against, so the answer flips: worth +$9.15M at the blended cost of finance, -$1.16M against what shareholders need (Fig. 4.1). In 79% of simulated scenarios it lost value (Fig. 4.3). The sharper finding was cash: for its first three years the plant can't earn enough to cover its loan repayments (Fig. 4.2). I recommended going ahead only with repayments shaped to match the plant's income and a cash reserve to cover the early shortfall.

A business case isn't finished when the numbers look positive. It's finished when someone has gone back hard enough to find what the first pass got wrong, and said so. That's what I'd bring to an investment paper.

An individual report comparing how well Australian government agencies run projects at federal, state and local level, alongside a group project on why Melbourne's upgrade of myki to tap-and-go bank card payments fell years behind London and Sydney.

I rated six agencies against a standard project-management maturity scale, two at each level of government (for example, the Tax Office and the National Disability Insurance Agency federally). For the group, I built the investment logic map for the myki contract (Fig. 5.1), and the international comparison behind the London, Sydney and Melbourne timeline.

Victoria's transport department ran a sound tender, but signed the contract with known schedule risks still unresolved. The result was an 18-month standstill and a $136.8M cost increase. The technology wasn't the problem; London and Sydney had already proven it. The failure was governance: the risks were written down, but never turned into a decision before signing.

This is the check I'd bring to a project office: don't sign off on an optimistic schedule, and know how to find the exact point where process stopped giving real control.

A ~2,870-word consultancy management report on the Woody 2010 case: a custom woodworking company's $17M capacity expansion that collapsed into cost overruns, schedule slippage, and unresolved supplier disputes.

I used the industry-standard project management framework (PMBOK) to check six areas: who had authority, stakeholder communication, scope, cost, schedule and close-out. I traced each failure back to the control that was missing and paired it with a fix (Fig. 6.1): a written scope with a formal process for approving changes, a plan for who hears what and when, tracking of progress against budget, and a close-out checklist.

The project hadn't failed from one bad decision; it had never had the basic controls to fail well: no charter, no work breakdown structure, no cost or schedule baseline, no stakeholder register, no close-out review. I delivered a full set of recommendations to rebuild those controls before the next capital project, not after the next failure.

This is the PMBOK toolkit applied end to end rather than module by module. It's the same lens I'd use to diagnose a struggling program: work backward from the symptom to the missing control, and give the fix an owner.

A group project for BUSM4764 (Agile Project Management), applying Agile and Lean to Netflix's in-house content production, where fragmented, stage-based workflows across creative, production and tech teams were driving delayed releases, missed audience trends, and higher cost and rework.

Used value-stream mapping and root-cause and bottleneck analysis to locate where the fragmentation actually lived (Fig. 7.1): sequential handoffs, low visibility and reactive late-stage decisions. I then designed the fix (Fig. 7.2): a shared Scrum cadence across creative, production and tech with regular reviews to surface issues earlier, supported by Lean and Kanban for waste elimination and real-time flow visibility.

A staged implementation roadmap: pilot one production team, establish sprints and retrospectives, bring external studios into shared visibility, then scale across content divisions. Success is measured on flow, delivery and collaboration metrics rather than on whether teams were simply "doing Agile." Backed by the ICAgile Certified Professional (ICP) certification I earned in the course.

A delivery methodology only earns its place if it's chosen for the bottleneck it fixes, not applied by default. This is that instinct: find the actual constraint first, then match the framework to it.