Turning unused dashboards into a decision engine for 100+ centres

A large education provider operating across Australia, New Zealand, and the United Kingdom had built a centralised, mature, comprehensive data environment – and almost nobody was using it. Centre managers were too buried in operations to open dashboards, and slow, manual reports meant the group was always acting late on student disengagement, tutor performance issues, and revenue opportunities.

They needed to go from "data exists" to "decisions happen," without expecting 100+ centre managers to learn SQL.

We built intelligence layers on top of the existing data environment: an Automated Insight Engine producing weekly "Top 3 Insights" per centre from a rules + ML approach; Predictive Models for student churn, performance decline, and revenue at risk; a Natural-Language Access layer answering manager questions as visual outputs in under ten seconds; and a Benchmarking Layer showing each centre against group benchmarks.

Tech stack: Databricks, Snowflake, BigQuery, Apache Airflow, MLflow, AWS. First value at Week 5, full capability at Week 9.

Manager engagement with data rose 3.3× and reporting load dropped 70%, freeing centre leaders to act on insights instead of assembling them.

Revenue per centre and retention both rose ~12%, and the programme hit 280% ROI in Year 1 – the data environment finally paid for itself.