Heathrow Airport builds a machine learning passenger flow forecasting model on Databricks

Overview

Heathrow Airport replaced a spreadsheet-based passenger forecasting process with an ML forecasting model on the Databricks Data + AI Platform. The new model cut forecast turnaround from two weeks (two people) to four hours (one person) and reduced the flight-level forecasting error margin from 30% to 10%, letting the airport proactively plan cleaning, maintenance and staffing around passenger wait times.

The challenge

Heathrow relied on over 500GB of data weekly and legacy, siloed data tools to power more than 1,300 daily flights; without distributed computational power its forecasting model required two weeks and two people to manage, and the airport lacked tools for user education, data governance, security and ML model training.

The solution

Heathrow centralized its data and analytics on the Databricks Data + AI Platform on Azure, starting with a passenger flow forecasting model, implemented data governance with Unity Catalog, trained users through Databricks Academy, and integrated Power BI with Databricks for dashboards and visualizations.

Reported business value

Heathrow sped forecast insights from two weeks and two people to four hours and one person while decreasing the margin of error from 30% to 10%, enabling proactive planning of predictive maintenance, cleaning and service interruptions around passenger wait times.