Leveraging AI to Predict Order Disruptions

‘Efficiency’ roots from processes, solutions, and people. It is the driving force shaping the way companies work. It is also the ingredient that propels productivity while allowing us to be environmentally friendly.

LafargeHolcim's most innovative series of concrete solutions -"Ready-Mix (RMX) Concrete" is designed to help companies improve their buildings’ energy efficiency, reduce costs, increase worksite productivity, and reduce carbon footprint. It provides higher quality, consistency, and reliability while maintaining excellent mechanical performance to meet the customers' most stringent needs.

Customers have leveraged our Concrete Direct app to order RMX products for their projects while relying on our world-class service to be certain of on-time delivery. This is possible because our supply chain model is straightforward. We prepare the order by date, call the driver the day before, and load the concrete the next morning. The driver delivers the exact concrete to the specified address.

A good percentage of customers go through last minute order changes that would result in operational losses due to cancellations close to / on the day of scheduled delivery. However, the truck drivers and concrete workers that had been pre-scheduled to service the order would call-in as per the original customer order with the need to be compensated.

We realized that there was a lack of a data driven approach for identifying orders that can get cancelled and that there was an absence of a prioritization methodology for labor planning to fulfill the orders. We wanted to build a solution to empower dispatch with a robust machine learning enabled solution to effectively get ahead of order cancellations and minimize excess labor expense. So, we developed predictive models to flag orders at risk of cancellation and simulated annual operational cost savings in several millions of dollars through our network.

One of the several reasons we see them canceling orders is the weather. After running the ML models we learnt some additional interesting facts such as:

• Big ticket orders are less prone to cancel compared to smaller orders
• Orders are more prone to get cancelled on a Monday compared to other weekdays.
• Customers and projects with a history of same day cancellations are prone to continue it.
• High cancellations till a day before scheduled delivery indicate possibility of more cancellations.
• Customers are less prepared for colder days during relatively warm months.
• Low temperatures and/or high precipitation lead to more cancellations.

We designed algorithms to flag such orders so that our dispatch team do not have to prepare for such deliveries.

To improve the demand and supply planning process's efficiency, we had to encounter the risks associated with the order cancellation. Although, we put forward various innovative ideas, the results did not reflect the expectations, resulting in the loss of thousands of drivers’ hours. We wanted to leverage a powerful AI-enabled solution to empower ‘order dispatching’ to effectively get ahead of order cancellation and minimize waste.

Roadmap that led to the solution’s development

The solution design team used retrospective data in the sterile concept (the idea was to solve as many challenges as possible for a Proof of Concept. With the field team giving positive feedback, the cloud-based working model was deployed with a real-time front-end provided to dispatchers.

The deployed solution has helped reduce not only the idle hours but also flag the cancellations that we usually would have missed with our heuristic models. It also has helped estimate market-specific dynamics and the road ahead is to roll the solution to all US markets.

Key findings and refinements in the algorithms

Labor planning is a holistic process

An effective labor plan must deliberate factors other than the quantity (orders), such as the distribution of orders throughout the day, the value of the relationship with customers, and so on. Therefore, the model output was modified to predict the quantity based on the hourly forecast.

Order quantity may vary with resource plan

‘Order quantity’ shows a considerable variation between the forward order book and the tickets, making it impossible to use it as a predictor variable.

Resources are reasonably fixed during the day

This contradicts human assumptions that resources will be concentrated in the market on a given day. This has led to corresponding changes in forecast reports, accuracy calculations, etc.

Ease of use

No matter how sophisticated the cancelation metrics are, the variables that impact the solution, levers that work, it is the adoption which holds the true value. The ease of use of a solution is what helps realize the potential of any digital initiative. The solution design team built an end-to-end solution comprising intuitive UI screens & functions, automated data flows, and model runs. So, all we had to do was measure the impact in business terms.

Impact

• Less waste for the planet and reduced costs that help the business be more efficient
• Streamlined on-time dispatch for better customer experience
• Saved idle hours % indexed for YoY growth in the RMX business
• The team’s satisfaction of making a positive change in the world and LafargeHolcim’s yet another step towards being more carbon efficient.

Note: LafargeHolcim’s digital and data teams worked closely with data scientists and engineers from Tredence on this initiative.