Disaster Forecast: How we can use predictive analytics to optimize transportation costs during event uncertainty.

Within any supply chain, the effects of adverse weather can have maladaptive consequences on the bottom line. However, when a company’s operations rely upon the forecasting of such weather events, a wrong prediction can have dire consequences. In the case of The Home Depot and its response to an oncoming hurricane, the company needed to procure supplies well in advance of the event, and distribute them across the four regions of the area. The challenge was two-pronged: 1) How much supply should The Home Depot initially procure, and 2) how should The Home Depot allocate these supplies amongst the four regions. If initial procurement were lower than demanded quantity, expedited shipping fees would have to be paid. If supplies were not properly distributed between regions, transportation costs would be incurred.

In our initial simulation, we opted to procure only the more expensive items and order the other ones at the expedited rate, distributing them using a weighted average ratio based on the likely path of a hurricane. Later on, we opted to procure items based on the lowest total average cost, which was derived from 20 different scenarios. Additionally, our distribution methods changed to follow a variance of percentage model, where we opted to distribute supplies based on the lowest predicted likely cost of transportation. Our final recommendations revolve around The Home Depot opting more for our second model while forging relationships with local suppliers and distributors to reduce overall costs. Furthermore, we recommend The Home Depot stick to its comparative advantage regarding supplies and opt for the local suppliers to meet the demand of general supplies in the event of a crisis.

Core challenges:

As a significant home improvement and repair supplier, The Home Depot (THD) is a uniquely positioned actor before, during and after natural disasters such as hurricanes. THD’s retail and logistics locations and workforce are themselves affected by the disruptions caused by hurricanes, while consumers rely on the company’s products as they prepare for and respond to storm damage.  The challenges THD faces in managing these duals roles include the following:

  • Predicting phases of customer demand during the hurricane season

THD faced difficulties in predicting the types and amounts of products demanded by consumers in storm situations. Ineffective forecasting efforts during Hurricane Andrew in 1992 resulted in both shortages of in-demand products as well as overages of products that went unsold and ultimately had to be salvaged.

  • Predicting the hurricane path to reduce reallocation costs

Given the impossibility of predicting the final landfall point of a hurricane, THD is unable to definitively determine the best geographic distribution of products. THD must, therefore, develop a system to distribute goods in a manner that optimizes transportation costs with regard to all hurricane path probabilities.

  • Establishing appropriate and ethical prices before and after the hurricane

In line with THD’s corporate values, the company established a “no-profit” policy during Hurricane Andrew to avoid price gouging during disaster situations. However, regardless of these price freezes, THD faces higher costs due to expedited pricing and transportation to supply its stores during hurricane situations. As such, proper forecasting is vital to ensure THD financial sustainability.

  • Managing and maintaining workforce levels before and after the hurricane

THD prioritizes the safety of its workforce in the event of a hurricane. However, communications with staff during a disaster situation are difficult to maintain, and THD’s operations rely on workforce resilience and capacity to serve its market.

  • Reputation and customer loyalty

Even as the operational realities of a company like THD become irregular during natural disasters, customers continue to build their perceptions of company reliability, product offerings, and customer service. Beyond THD’s direct revenue gain or loss during the storm situation itself, company reputation has a lasting effect on company success.

Lessons learned:

  • Allocation

Initially, we attempted to manage costs by allocating resources based on the weighted average of the probable demand of all three scenarios. This seemed to be a sound choice, as when one considers the long-term effects of disaster management, a weighted average would lower overall costs in the long run. However, in our initial response, we stated that our allocation was based on a onetime event. Thus we wanted to allocate funding based on that event, and avoid any drastic fluctuations in price. That is, while the weighted average model would be ideal for multiple scenarios over time, we opted to reduce the volatility risk for a one-time event. However, after analyzing the shipment costs in addition to the expedited costs using the procurement model, we opted for a different allocation method, where we derived the numbers from a variance of percentage model (see appendix 1). Using this model, rather than calculating price based on a per item basis, we calculated the cost of moving one percent of the supplies between the regions relative to one another, e.g., the cost of moving 1% from region 3 to 1, or region 2 to 4. The percentage of variance calculations allowed us to better determine the high-cost regional shipment pairs and gives us a better conceptual understanding of where to allocate the supplies efficiently.

  • Procurement Practices

At the beginning of the game, we worked on a procurement strategy that focused on delivering 75% probable demand for all the products most profitably by only ordering discounted items that are too expensive to procure expedited. We did this by optimizing the solution for a minimum value of combined discounted ordering and expedited ordering. Additionally, we opted not to order water bottles, as we determined the cost of expediting water during times of crisis would, in fact, be cheaper than the overall cost of moving the water between regions due to shortfalls.

Upon further analysis, we realized the problem with this approach: while it considers the combination of procuring costs involved, it doesn’t consider the costs of shortage and overage. The shortage and overage costs are important, as we are looking to reduce costs to the lowest 10th percentile, and these costs make up a considerable portion of the overall supply side costs. Another problem was assuming similar demand for all products; demand is more random and dependant on variables such as the extent of damage, power infrastructure, water, sanitation conditions, etc. So it is idyllic to consider various scenarios of demand for these products and model a solution which is optimal for any of these conditions and consider the overall supply costs, not just the procurement costs.


As outlined through the lessons learned, there are a variety of ways to calculate the ‘optimal’ scenario. However, due to the sheer quantity of possible variations from the challenges outlined above, we opted to focus on a way to minimize overall costs by focusing on average cost increases, rather than try to mitigate the worst case scenario. One of the goals THD should focus on is its ability to provide those services which are its comparative advantage as a store. Thus regarding allocation, items such as bottled water should not be prioritized, considering they could be acquired from other retail stores.

On a macro level, another way THD can potentially mitigate some of its supply chain risks is by ensuring synergies exist between the command center and stores to promote coordination with the regional responses. Also, communication with other large-scale retail stores can be beneficial as there will be an industry-level need to improve logistics and can serve as a backup option in case there is any hold up with the current THD recovery plan. We recommend that THD continue its socially responsible, price stabilization policy to ensure it can provide to its customers in times of need. Finally, it is essential to ensure that their distribution centers have a more stable structure in place to be able to provide to the stores. In the future, THD can consider adding an additional distribution center, if it will significantly lower the expedited costs and provide more supplies when in need for a post-hurricane strategy.