Predicting the unpredictable

This year’s peak period will be the largest in Australia’s history. Last year Australia Post delivered more than 40 million parcels in December, and this year, volumes are already greater than they were at the same time last year.

Over the coming peak Australia Post expects to deliver up to 3.5 million parcels on its busiest day- this level of demand puts a lot of pressure on logistics providers to deliver.

For Jhordan Gil, senior solutions advisor – Supply Chain Execution at JDA Software, organisations should start to explore artificial intelligence and machine learning to more accurately predict peaks in demand at a very localised level and to better prioritise warehouse tasks and position assets to deliver on those peaks.

According to Jhordan, the real benefits from artificial intelligence and machine learning come when logistics providers or retailers have to cope with unpredictable peaks in demand.

“With the amount of data and technological advancements available now, we can begin to predict the previously unpredictable,” Gil said.

For example, if there is a football match close to a Coles or Woolworths, and the forecast calls for the first warm and sunny weekend in the last few weeks, an AI/ML based demand forecasting and autonomous fulfilment system may increase stock at that branch and correspondingly schedule additional staff at that particular time. It’s these kinds of spikes in localised demand at scale that were difficult to prepare for without advanced technology about historical behaviour and patterns.

A further advantage of using technology is that it gets better over time, Gil said. “It starts to recognise a good outcome or a bad outcome, and adapt recommendations accordingly.”

A crucial aspect of coping with spikes in demands is having the right skills, in the right place at the right time, Jhordan says. “When you identify a peak period, it has a huge effect on your workforce. So, it’s advantageous to be able to tap into a diverse range of employees, understanding which functions of your operations will be most affected, and select the right profile of employee you need at any one time,” Gil said.

With regards to logistics providers, one piece of advice that Jhordan offers is to improve collaboration. “For logistics organisations, nearly everything is out of their control. We recommend that providers work with their clients to improve collaboration and traceability. The idea is to ensure that everyone sees a single source of truth,” he says.

Whether it be a container in the middle of the ocean, a truck on the road, or the last-mile delivery van, it pays to have accurate ETA data constantly available to all parties involved in the supply chain. Further, AI/ML can be leveraged to go from Estimated Time of Arrival to AI/ML-based Predicted Times of Arrival, allowing organisations to better adapt depending on if a delivery is running late or early.

Within warehouse environments, Gil encourages following the principle of having the right task at the right time rather than simply 100 per cent utilisation. Delivering on time in full (OTIF) becomes particularly challenging during peak periods, and AI/ML can be leveraged to intelligently assign warehouse work to the right employee at the right time. For instance, if a set of urgent orders drops, how do you ensure you can redeploy staff so they can complete the urgent orders without affecting service level of other orders? Live accurate data enables you to prioritise your assets and execute on the unpredictable accordingly,” he says.

JDA’s Luminate platform embraces SaaS-native architecture, the Internet of Things (IoT), and advanced analytics and cross-platform integration with AI and ML to drive value for customers’ supply chains.

JDA offers a control tower which senses unexpected events across the supply chain and identifies the potential impact of disruptions. Analysing hundreds of potential variables to find the probabilities of different outcomes and each outcome can then be associated with a calculated business impact and risk.

“These technologies aren’t just applicable to forecasting unexpected peaks, they also really come into play once you’ve hit that peak period and you need to work out how to deliver,” Gil said.

For Gil, it’s about having that central source of truth and having the smart software to recognise what needs to be done when unexpected changes occur.