Operations portal for a regional Victorian utility

The context

The client is a regional Victorian utility with a small corporate team and a much larger distributed field workforce. Day-to-day operations were tracked across four shared spreadsheets, a legacy FileMaker database dating from the mid-2000s, a SharePoint site nobody trusted, and \u2014 in practice \u2014 a long-running WhatsApp group.

Nothing was broken, exactly. It was just slow. New starters took six weeks to become productive. Asset data drifted silently. Reporting for the board was a monthly overnight reconciliation by the operations manager.

What we did

We ran a four-week discovery before writing any code. We spent time with field crews, administrators and the operations manager. We read shift handovers and sat in on incident reviews.

The resulting brief was blunt: a single operations portal, with clear permissioning, fast search, auditable changes and asset-level history. No replacement for FileMaker — a replacement for the need to know FileMaker.

The build

• Stack: Astro for the marketing skin, an Inertia/Laravel application for the authenticated experience, Postgres for data, Cloudflare for the edge.
• Auth: single sign-on via Microsoft 365, with role-based access and break-glass admin accounts documented in the runbook.
• Imports: a robust ETL for the FileMaker export and CSV attachments from the spreadsheets, with dry-run, diff and rollback.
• Mobile-first: the portal had to work at a depot, in a ute, at 6am. Performance budget measured per-view.
• Observability: OpenTelemetry traces from day one; error budgets agreed with the operations manager, not just engineering.

Outcomes

• Onboarding for a new field starter reduced from six weeks to eight working days.
• Monthly board reporting reduced from a three-hour reconciliation to an exported PDF run by the operations manager herself.
• Audit trail now covers 100% of asset updates, with change history attributable to a user and a device.
• FileMaker retired on schedule. The WhatsApp group still exists \u2014 for jokes.

What we\u2019d do differently

We originally planned three rounds of user testing; we ran five. The additional two rounds caught a permissions edge case that would have caused real harm in production. If we were starting again, we\u2019d budget for five from the outset.