As volume of data continues to grow at a staggering pace, manufacturing organizations across the globe are using multitude of methods to draw insights from their data (internal or external). The focus on “data as the new oil” can be seen in their efforts to govern their data (cleansing of garbage data), management of the data (through data lakehouse) and through the usage of business intelligence and data science solutions.
Data science (AI / ML) are used across the manufacturing value chain from production improvement to quality improvement and to cost optimization. In the present case, we will focus on certain cost optimization solutions for a steel manufacturing organization.
For the steel major, once the finished goods are produced in the production facility, they are moved to the stockyard from where they are picked for final dispatch to the customer location. During this process, goods are loaded and unloaded multiple times requiring considerable time in terms of effort and fleet. The requirement was to streamline the process by optimizing the truck load, forecasting the freight rates, and monitoring the vendor performance.
Truck load optimization:
The client needed to identify the possible number of trucks that are required in a month for a particular customer for a specific destination and to identify availability of trucks that are ready to ship/ dispatch on a daily basis for all available customers.
A constraint based linear programming was formulated for minimization of total number of trucks/trailers required for a particular customer for a destination. Customer demand, average computed cost by vendors(min), vehicle capacity specific to given sales order number and material code were used as constraints.This allowed the system to throw back the number of trucks and the load that they will need to carry from the production facility /stockyard to the customer location.The automated solution has provided an overall reduction of 96% effort to create the dispatch order generation process.
Freight Rate Forecasting:
The company, as one of the leaders in their field, provides their services all over India. Their need to transport and the complexity of the trucking market leads to tremendous uncertainty leading to volatility in the rates they would need to use. The organization wanted to forecast the freight rate of any destination so that it’s related plans or actions could be strategized accordingly.
Using the dispatch report of past 18 months for different destinations information, we defined 6 different zones i.e., East, West, South, North, Central and North-East, in India.Basis the defined zones and its respective destinations, different clusters were formed basis latitude and longitude and distance from each other using hierarchical clustering algorithm. This information was considered as a variable that was then used for developing a regression model for each zone.
The regression models were developed with clusters and vehicle type as categorical variables and the month index was used as a numerical variable for all defined zones. These models helped to predict the forward freight rates for any of the new / existing destinations using the cluster characteristics with accuracy levels of more than 85%.
Vendor Performance Management:
Using the consolidated dispatch report and the accident master sheet, 40+ performance indicators were defined for all vendors based on % deviation between the actual and expected vehicles placed and number of accidents in the past faced by any vendor.Each indicator was then scored (on a 5-point scale, 5 being best) across all vendors by labelling the unique values.
Indicators were then ranked with respect to their importance in the manufacturing setup and a weightage was assigned to each indicator based on the rank.The weighted sum of the scores across all indicators was then used to monitor the performance of the vendors on a real time basis leading to increase in efficiency.
GPS Tracking Dashboard:
The developed solution enabled the company to track the movement of the vehicles where GPS was installed on the vehicles. The solution has been integrated with the vendor site to extract information using API for all the open trips and the visualization tool reflects the real time movement of the vehicles along with key important metrics like idle time, stoppage time etc. This solution provides granular level details of tracking of the vehicles leading to better management of the delivery schedule.
Dhrubabrata Ghosh is the Managing Director, Data & Digital, Protiviti Member Firm for India.