PARIS – The air cargo industry often touts digitalization, technology and data as tools to alleviate pain points challenging operations. Even as the industry shifts to adopt these tools, data as the crux of these tools must also be reconsidered. Namely, data must be collected and leveraged more intelligently, which when combined with machine learning can serve to create a “crystal ball” for companies.
At the Air Cargo Handling Logistics conference in Paris (CDG) today, Hermes Logistics Technology CTO Alex Labonne explained how companies in air cargo can develop their own crystal balls.
Machine learning is one way companies can develop foresight into their operations, by utilizing data inputs and outputs to predict future costs, loss and revenue, as well as detect problems and model potential new business or markets segments. However, data must be well prepared to be leveraged to its highest potential.
“Data today is becoming more valuable than oil, but like oil, data needs to be extracted and refined to be consumed,” said Labonne. “When you begin to think of data no longer as a set piece of information, but as an event, everything changes,” he continued.
Events data includes not only the present set of information related to a cargo shipment, but also includes historical data related to the shipment and its changes over time. Industries like fintech and banking are already leveraging events data through machine learning, but the air cargo industry has little used this massive source of determining business potential.
The main roadblocks to the air cargo industry adopting the technology and data collection methods, Labonne said, are predominantly costs, security and the way the industry thinks about data. Though entry costs may be a challenge, they are nowhere near the amount they were ten years ago, and the technology needed for these processes is readily available online to companies today. Machine learning also does not suffer from encrypting or encoding information, so companies can do that if they like. Rather, the biggest issue is that companies much change their mindset toward data.
Some companies may not collect historical data or elements related to the change of a shipment over time. If they do not, the easiest solution then is to add it into the data you do collect. Companies can design databases as they like and are not required to throw away their existing systems, but to correct the trajectory of data collection to treat data as events and make sense of the old information that may otherwise have been thrown away.
By redesigning databases to serve machine learning through event data collection, companies acquire context and meaning from past events which can serve to create new insights and in turn, business.