How Transport for NSW lowered data-lake OpEx by 22%.

Transport for NSW operates the digital backbone for buses, trains, ferries, and roads across New South Wales — a real-time transport platform serving millions of journeys per day. The platform's data lake on Azure handles billions of rows of telemetry, journey, and ticketing data.

In 2023, the data platform team faced a problem most public-sector cloud teams will recognize: spend was growing faster than the workloads it was supporting, and finance was asking questions the team couldn't answer in less than a week.

The challenge

TfNSW's data platform combined Azure Data Factory, Synapse Analytics, ADLS Gen2, and a cluster of Azure SQL databases. Spend was distributed across multiple subscriptions with inconsistent tagging — making per-team allocation manual and error-prone.

Worse, the spend trend was clearly going one way. Without an intervention, the data-lake OpEx line was forecast to exceed budget by mid-2024.

The CloudMonitor deployment

TfNSW installed CloudMonitor inside their tenant on a Wednesday afternoon. By Friday morning, the first $40k/year recommendation was already in the queue — a misconfigured ADLS lifecycle policy that was keeping cold data in hot storage indefinitely.

Three changes that drove most of the saving.

1. Storage tiering on ADLS. CloudMonitor surfaced ~32 TB of telemetry data that hadn't been accessed in 90+ days. Moving it to cool tier and adopting a 180-day archive policy saved $14k/month.

2. SQL DB right-sizing. Half the SQL databases were running Business Critical when General Purpose would have met the SLA. CloudMonitor's right-sizing recommendation queued the move; engineering applied it over a sprint.

3. Reserved Instances on Synapse compute. CloudMonitor's RI analyzer identified a 1-year RI pattern that matched 89% of historical Synapse usage. The team committed and captured 41% on those workloads.

Outcomes

Twelve months on, TfNSW's data platform OpEx is 22% below its baseline, despite supporting 18% more daily workload. The CloudMonitor reports are embedded in every quarterly review, and the cost-group view is now the canonical reference when engineers make architecture decisions.