The auto industry serves as a perfect exemplar of three key eras of statistics and data science in service of industry:
Total Quality Management (TQM)
First in Japan, and later in the U.S., the auto industry became an enthusiastic adherent to the Total Quality Management philosophy. Fundamentally, TQM is all about using data to improve processes, and, hence, the outcomes of those processes. Two contributions from statistics lie at its core.
- Statistical Process Control (SPC), which draws a line between random normal variation in a process, and unusual deviations that require attention. SPC helps companies avoid counterproductive and destabilizing meddling in processes in response to what is nothingmore than statistical noise.
- Design of Experiments (DOE), which adapts the principles of statistical studies to the industrial setting. Multiple factors are involved in most processes, and testing all of them at different levels for standard group comparisons is prohibitively expensive and time-consuming. DOE theorists and practitioners have built a rich set of experiment designs (and corresponding analyses) that extract maximum information from a minimum number of experiment runs.
Big Data Analytics
The explosion in data generation and availability allowed predictive analytics to be applied across a wide spectrum in the auto industry, from manufacturing to servicing to sales and marketing. Models can dig through reams of data to help identify faults before they become major safety hazards, by identifying the signal amidst the data noise sooner than would be possible through simple observation. Hence, most auto recalls are proactive in nature, and based on relatively few hazardous incidents.
The use of big data analytics in this phase is in keeping with the trend of continuous improvement – its contribution is evolutionary, not revolutionary.
Artificial Intelligence (AI)
Nowhere has the application of artificial intelligence received more attention than with self-driving cars. With sensors like radar and sonar, multiple cameras, GPS and more, a self-driving car may deal with a flow of more than 4,000 gigabytes of data per day.
A bevy of rules-based decisions systems coupled with image recognition and predictive algorithms governs the car as it drives itself. These statistical and machine learning algorithms are not in pursuit of incremental change, but rather a revolutionary transformation in the auto industry. A “moonshot,” in the words of Tom Davenport.