Volkswagen Data Analytics Project

I completed this project as a student of the QUEST Honors Program which is a multidisciplinary, hands-on program for STEM students that focuses on quality management, process improvement, and system design through teamwork and co-curricular programming with corporate sponsorship.

As part of a capstone project, I collaborated with a QUEST team for 3 months to use data analytics to create a streamlined process for car delivery using a DMAIC methodology. By interviewing stakeholders at the port and dealership, the team was able to identify wastes in current information flow and port layout. These wastes create delays across the port that prevent port workers from being proactive. Data was first accessed from MS SQL server and cleansed for use by the analytics model. Weidentified importantvariables in predicting the amount of time it takes a car to get to the dealership. The prediction model was built using R and the results were presented using Lucidchart. By analyzing the physical process, the team was able to eliminate wastes and make it easier for Volkswagen to accurately estimate car delivery times. 

Problem:

  • Since its founding in 1955, Volkswagen Group of America has expanded to over 8,000 employees and 1,000 dedicated dealerships, gaining a reputation for excellent customer and vehicle service. Part of this service includes providing both dealerships and customers with a vehicle delivery estimate for their specially ordered vehicles; however, the accuracy of this system ranged from 20% to 55% on a month to month basis.

  • The team was initially charged with developing an improved model for predicting delivery estimates

  • First, the team sought to locate and eliminate sources of lean waste within the delivery process to reduce overall variation that may be difficult to predict for. 

  • Next, the team sought to develop an improved model by determining the most significant variables affecting delivery times.

  • The purpose of this approach was to identify and eliminate process variation that would contribute to accuracy.

Research Approach:

  • First, the team sought to locate and eliminate sources of lean waste within the delivery process to reduce overall variation that may be difficult to predict for. 

  • Next, the team sought to develop an improved model by determining the most significant variables affecting delivery times.

  • The purpose of this approach was to identify and eliminate process variation that would contribute to accuracy within the new model.

Analysis Approach:

  • Conducted quantitative analysis on 2018 vehicle delivery data using Excel, SQL and R.

  • Performed qualitative interviews of stakeholders and analysis at the Port of Houston and Audi Silver Spring

ANALYSIS Results:

  • Upon initial analysis of average times and variations between each step in the process, it was found that the steps with the highest variation were the ones involving transportation.

  • Further regression analysis revealed that the vehicle brand, month in which the vehicle was allocated, type of transportation used, and the dealership location were the best predictors of delivery times.

  •  The dealership staff asserted that they and the customers have limited access to information about the delivery process. Their dashboard provides the delivery estimates and the most recent checkpoint; however, they have no access to a car’s status in between checkpoints. Therefore, if they know a car was released to carrier but need more information, the dealership has to call a help desk who in turn reaches out to the carriers. These barriers to communication have resulted in frustrations for both the dealerships and their customers.

Recommendation:

The team developed the following three recommendations:

  1. The creation of a proprietary software for facilitating communication between Volkswagen and third party vendors

    • With 5 different ports and multiple different carriers, there are at least 15 different systems at play in the United States, most of which do not communicate with each other. By having one centralized system, all parties will be able to access the information when they want it and how they want it.

2. The restructuring of port layout to accommodate the increasing throughput requirements

  • Change current functional layout to a cellular layout

  • Cellular layout allows the vehicles to be in proximity to multiple work locations and can be taken to the nearest one. This creates a more linear flow of vehicles and reduces the amount of time wasted through movement and transportation.

3. The implementation of an improved delivery model based around significant variables and historical data

  • Volkswagen’s original delivery estimate was based solely on the previous month’s performance

  • In contrast to the original model, the new model incorporates annual historical data as opposed to month-to-month predictions. This approach serves to normalize non-recurring events and to account for the seasonality of vehicle delivery.