The company’s continuous growth in sales revealed some issues associated with their delivery model. The KML regions, which were manually defined, did not dynamically adapt to spikes in demand, driver emergencies, or sales trends in each region. As a result, Hyperwolf was experiencing inventory issues and extended delivery times. For companies like Hyperwolf, optimization of resources is critical to achieving efficient and effective business operations while providing great service to customers. In order to optimize operations, drive business growth, and enhance profitability, Hyperwolf recognized the need for dynamically generated KML region files and decided to partner with Data Sleek.
After looking at different options, Data Sleek decided on a multistep approach for generating the dynamic and optimized geographic data file HyperWolf needed:
1. Build a data warehouse
2. Pre-process the dataset using features relevant to the problem
3. Geographically cluster historic data on deliveries using the K-Means clustering method
4. Utilize a geospatial data package to allocate clustered regions on a grid map, forming new KML regions
5. Output the desired combined KML file based on work shift, number of drivers, and day of week
6. Using Prophet, generate a demand forecast for each region in the KML file for a better business overview
The Data Sleek team introduced a forecasting method for predicting the number of sales. This has enabled both business analytics and performance overviews. The tool starts with the newly produced KML region file and a pre-processed dataset representing historic sales then outputs an order forecast dataset for up to 3 weeks in advance for each region in the KML file. Because the available dataset consisted of only a few months of historic data and business leaders conveyed an understanding that sales numbers were unstable in nature, the team decided to implement the forecasting tool for this problem using an existing forecasting package called Prophet. With some tuning, Prophet is capable of building and garnering fast and fairly accurate forecasting results at scale for all the regions.
The team at Hyperwolf now has an automated process to create delivery KML files dynamically. These files are tailored specifically toward their delivery pattern and customer demography. This solution frees HyperWolf from the time-consuming manual process of drawing KML regions and revolutionizes their business model’s operations with a continuous stream of automatically generated KML regions that are optimized to satisfy the end customers’ demands. In addition, the process helped to reduce the total number of drivers needed, streamlining operations and lowering a major business expenditure in staffing.