Heathrow Airport is using a range of data and analytics tools from Microsoft and machine learning models to better predict passenger flows through its terminals to improve operations and make passenger journeys smoother.
The key for Heathrow is going from a paper-based, reactive model for operations to a more predictive, proactive planning model where there are fewer surprises for staff on a day-to-day basis. Heathrow is one of the busiest airports in the world, processing nearly 215,000 people a day.
"The key is not reacting now but being predictive, so staff can change their plans," Heathrow CIO Stuart Birrell told Computerworld UK.
Back in 2016 Heathrow turned to Microsoft as part of a wide-ranging data modernisation programme which involved using Azure cloud services, including Data Lake Analytics, Stream Analytics, and Azure SQL Database, to extract, clean, and prepare real-time data about flight movement, passenger transfers, security queues, and immigration queues. They added Microsoft Power BI as a frontend analytics tool for the visual consumption of real-time metrics coming out of those systems.
"We went out to market and looked at four or five players for the platform and the analytics tool on top," Birrell explained. Where Microsoft came out on top was its ability to do both.
"We chose Microsoft as the integrated solution for this...it did help to have front and backend in one place and the capability when both are working together is a pretty seamless integration, from the identity and security layers to data security and integrity."
"The big Azure challenge was getting that data from legacy systems into that normalised environment. Once we had that the rest came on pretty rapidly,"Â he added.
By the beginning of 2017 Heathrow had that data in front of security staff and by August it was speaking publicly about the results.
"We now have a machine learning model running to predict accurately the passenger flows by 15-minute increments into each terminal for managers to plan breaks and security lanes and schedule shifts and to balance that across the airport around peaks," Birrell said.
Say a bunch of flights come into the airport an hour early because of the Atlantic jet stream. Previously this would have forced immigration and baggage staff to scramble to react to the sudden spike in volume.
Now, thanks to this set of predictive models and the wealth of weather data available to the airport, Heathrow is able to share this insight with air traffic control and border staff to "create plans around that," Birrell said. "Historically you would have to stand and wait so that whole stress is reduced dramatically."
The next step for Heathrow now the data foundation is in place is to democratise the metrics it gives to frontline staff in different areas of the airport and to lean more heavily on machine learning to hone those predictive models.
Power BI reports and dashboards are already available to security officers, transfers and customer services staff. The next stage of the rollout will see border force given access to dynamic information relating to arrivals and baggage volumes throughout the day.
"It's very easy to apply those new lenses now," Birrell said. No longer is it the role of IT to produce paper reports 48 hours in advance, rather "we help with consolidating the data on Azure and have a data dictionary for common standards. We don't write reports and you now have people that understand data and analytics around the business."
Birrell also sees a huge opportunity to enrich the airport's machine learning models with new data streams. So for baggage handling, by "taking baggage data and aircraft departure times or volumes impacts the quantities of baggage, which has a major impact on performance," he said.
"So we can see if it is a holiday flight vs a Frankfurt shuttle, and by optimising that with machine learning we can find those patterns. It is early days but we are busy starting to experiment on that front."