From a macro perspective, it is evident that technology has helped supply chains overcome a variety of inefficiencies drawn from market fragmentation and a general lack of visibility and transparency into operations. That said, the industry still has to tackle a myriad of deficiencies within the system, as logistics is now a prime differentiator in the way businesses interact with their end consumers.
This shift in perception of putting consumers on top of the logistics pyramid is leading supply chains to gravitate toward business-to-consumer (B2C) models rather than continuing with traditional business-to-business (B2B) models. But for companies to expedite delivery, it is critical for them to understand order patterns, which are notoriously volatile due to steadily rising volumes and rapidly evolving consumer trends.
“Traditionally, order forecasts were done by looking at a historical stream of orders and using regression analysis to predict how the future looks like,” said Madhav Durbha, the group vice president of industry strategy at LLamasoft, an enterprise supply chain solution provider. “Now we are seeing an increasing interest in bringing additional data elements to improve the accuracy of that prediction.”
Understanding data streams will lead companies to discern patterns at a more granular level, helping them make forecasts based on artificial intelligence and machine-learning algorithms — techniques that rise above the limitations of traditional statistical models.
“We see this quite a lot amongst large online retailers who have order patterns that are extremely unpredictable due to high volatility. This sometimes causes brand owners working with these retailers to stop filling orders without thinking about the impact on the customers they need to service,” said Durbha. “This leads them to be hit with some punitive penalties. Being able to manage all that and still bring more predictability into volatile order patterns is the challenge that needs to be addressed.”
To create an effective forecasting model, companies need to start with gathering data and consolidate all the enterprise data into a single database. Noise in the data channels would also have to be weeded out to make sure the resulting data is good enough for rendering trends.
Durbha contended that for companies that sit on massive data piles, developing their analytics capability is a compelling investment. Once they gain insights into order patterns, it can be used by decision-makers across various levels of the supply chain — democratizing intelligence and making logistics a more seamless affair.
The rise of cloud and algorithmic intelligence has led to supply chains being digitally rendered, helping businesses to simulate real-world events and predict outcomes. Called “digital twins,” these virtual platforms provide speed and scale to companies, enabling them to plan for rapid market shifts.
“You need a system of reference from where you start connecting these dots and digital twins will start to emerge. Using that, you’ll be able to visualize information and be able to spot problems better,” said Durbha. “However, digital twins don’t show up overnight as you need to build the discipline organically.”
Ideally, a digital twin should allow companies to travel in time virtually to understand all sorts of possibilities and act on insights they gather. For instance, it can help with understanding replenishment frequency, choosing transit routes and even adjust for volatility in fuel costs.
“I see these inefficiencies to create tremendous opportunities for supply chain professionals, as there are several applications for automation and advanced analytics,” said Durbha. “From an organizational perspective, I think companies need to pay close attention to the idea of digital twins and leverage that to improve their operations.”