Australia Post looks to graph database, digital twin to improve delivery

Australia Post has reinvigorated plans for a digital twin of its delivery network, revealing an investment in graph database technology to make more predictive and proactive decisions based on millions of data points it collects each week.

Gary Starr.

Executive general manager of parcel, post and ecommerce services Gary Starr told the Online Retailer Conference in Sydney that Australia Post is “an early adopter of graph [database] technology”.

“This allows us to map our entire network digitally,” Starr said.

“In essence, it’s like a super database with real-time information allowing us to understand the relationships of connections in much more detail than traditional databases. 

“This is a step up from where we are today.”

Australia Post first flagged work on a digital twin of its delivery network at the start of 2021; however, the executive spearheading that work has since left.

It has also previously discussed the build of an event management system, based on Google Cloud Platform components.

It’s unclear if what Starr discussed is the same work, or an evolved version of these earlier works, since they now date back several years.

Certainly, Starr’s discussion of a “digital map” of the delivery network is the first time Australia Post has discussed these efforts in any great detail since.

Underpinned by graph database technology - vendor unknown at this point - the digital map of the delivery network will “look at nodes and the flows between nodes such as specific sites and docks”.

“From there, we can better identify bottlenecks, looping parcels and what we call ‘dark parcels’ where we lose sight of a parcel,” Starr said.

Starr is hoping to make use of “billions” of mail article movements across the delivery network every year - “millions” a week - to feed the graph database and digital twin.

“We’ll have what is effectively a live ‘super database’ of real-time events and be able to build a digital twin of our network to help model the best way to minimise and resolve disruptions,” he said.

“It also means we can model, when we add a new facility, how it impacts the network, [and] we can understand where to put an investment, how parcels will flow, what it will impact speed and ultimately the customer service in delivering that item. 

“It also means we can understand how particular modes of transport or rerouting items to alternate sites impact on delivery.”

On that last point, Starr said Australia Post already had some existing predictive capabilities enabled by investments in machine learning.

“Our machine learning is helping to power ecommerce by providing highly accurate volume forecasts, ensuring that our network, in partnership with our customers, can manage predicted volumes,” he explained.

“Volumes change dynamically and they’re impacted by what’s happening with retailers, but also there are environmental conditions, and with a network that spans the continent, we have to understand how floods, bushfires, accidents, derailments, a range of other events, impact on performance.

“Every peak period we say, ‘It won’t happen again’, and every peak period there’s some sort of major natural incident. 

“Great forecasting means we can help our customers plan for their volumes and we can work with them to understand those volumes so we can plan our network. 

“So, assisted by machine learning, we now have a forecasting product suite that features improved models to better understand and predict the likely path a parcel will take. 

“This is a model that looks at movements in the network based on lodgement location, destination postcode and article feature, and this is especially useful during our peak period.”

Starr indicated the graph database and digital twin would considerably improve Australia Post’s technical capabilities in this area.

“We’ll be able to understand how to minimise disruptions and work around a disruption much more quickly in a digital way,” he said.

“Investments in the digital map are really about supercharging how we use our data, moving from descriptive to investigative; highlighting congestion to alert our teams; resolving problems before they arise; diagnosing disruptions, and we can then figure out how to optimise delivery. 

“And for our customers, it means we can move to a much more proactive engagement so we can both get ahead of an issue.”

Building on strong foundations

More broadly from a technology strategy perspective, Starr indicated that Australia Post is continuing to invest in its data capabilities and culture.

“Historically we’ve been very data rich but insight poor,” he said.

“We know the difference and we’re making the right investments to ensure we’re an informed business.”

The “right investments” included establishing a data platform based on Google Cloud Platform, which dates back as far as 2019.

“In the last few years we’ve engaged GCP as our data platform, enabling us to be future-ready with the tech and infrastructure provided by Google,” Starr said.

“We capture billions of data points a year across our business, and we can use this data to draw out valuable insights.”

Starr said that Australia Post recognised that its future is “data-led”. 

While technology is one element needed to enable that future, building the right governance practices and culture is also required.

“What’s critical here is that we need clean data,” he said.

“Clean data feeds great decision-making.

“We’ve got to get the governance right. If we don’t get the governance right, how we manage data, and how we continue to protect our data as it evolves, then we just won’t get where we need to get.

“We had an excellent start to establish an excellent platform to pull all the data into one place, but we’re going to have to continue to build on that so we get better insights.

“That means really empowering our teams to be accountable for data quality, new tools, processes, frameworks, and metrics, so we build a stronger data culture so that every team is fully aligned with this.”