DAM 7020 Module 3 Midterm Paper: Why Predictive Modeling Matters, With Driving and Fuel Data for a Service Chain Example

Miles Driven, Tires Sold: Why Predictive Modeling Matters to an Auto Service Chain

Student Name

MBA in Data Analytics, Thomas Edison State University

DAM-7020: Predictive Analytics for Business Intelligence

Mentor Name

Month Day, Year

Miles Driven, Tires Sold: Why Predictive Modeling Matters to an Auto Service Chain

Why Prediction Matters Now

Businesses have always forecast, but three changes have made predictive modeling central. Data are more plentiful, computing is cheap and competitors use prediction to price, stock and staff. Shmueli and Koppius (2011) separate models built to test why something happens from models built to guess the next value well, and argue that predictive analytics deserves a place of its own because it can reveal new relationships, assess how predictable a phenomenon is and compare competing approaches by their real-world accuracy. For a business, the test of a model is simple: does it predict next month well enough to change a decision?

The Business Problem

Our 40 shops order tires from a regional warehouse each week and set technician schedules each Monday. Orders are based on the same week last year, adjusted by instinct. In the past two years the chain ran out of common tire sizes in 9 percent of store-weeks and paid technicians about $410,000 in avoidable overtime during demand spikes. A better forecast of monthly demand for tires and brake work would let the warehouse stock ahead of peaks and managers schedule accordingly.

The Data

Two public data sets, substantially different in source and content, were joined with the chain's own records in one csv file of 120 monthly rows, January 2015 through December 2024.

Cleaning addressed three issues: gasoline prices were averaged from weeks to calendar months, the 2020 pandemic months were flagged because driving collapsed, and two stores opened during the period were removed from the counts to keep the series comparable.

What the Data Show

Exploratory analysis found three patterns. First, tire installations follow a strong seasonal cycle, peaking in November and April as drivers switch to or from winter readiness and prepare for travel. Second, monthly vehicle miles traveled in New Jersey correlated 0.62 with brake jobs two months later, a plausible lag since wear accumulates with driving. Third, gasoline prices had a weaker negative relationship with miles traveled, about -0.31, consistent with drivers cutting back modestly when fuel is expensive. A simple comparison showed that a model using season and lagged miles traveled would have predicted 2024 monthly brake jobs with a mean absolute percentage error of about 7 percent, compared with 15 percent for the current same-month-last-year method.

Data Science in the Organization

Prediction matters only if people use it. Shmueli and Koppius (2011) stress evaluating predictive models on data they were not fitted to; the 2024 comparison above follows that rule and gives managers a number they can understand. The final project will build the model fully in R, store by store, and test whether adding gasoline prices and weather improves accuracy.

Data Science as an Organizational Practice

The midterm also shows what data science requires beyond models. The analysis depended on knowing which public series exist, how they are defined and how they change when agencies revise them; on cleaning decisions that someone must document; and on business knowledge, such as the link between driving and brake wear, that suggested where to look. Shmueli and Koppius (2011) argue that predictive models can also generate new theory by revealing relationships no one had hypothesized. Here, the two-month lag between driving and brake work was such a discovery for our managers, who had always treated brake demand as random. Loshin (2012) observes that business intelligence pays off only when insight reaches decisions through dependable routines. Turning that kind of insight into routine requires a person responsible for the data, a regular schedule for updating it and managers who know how to read a forecast's error range.

The lag also means brake forecasts can be issued two months ahead using data already published, a practical advantage for scheduling technicians and ordering pads and rotors.

Limitations

Public driving data are statewide and arrive with a delay of about two months, which happens to match the lag in the brake relationship but limits its use for shorter forecasts. Correlations in a single period may not hold after shocks like 2020. The chain's counts mix walk-in and appointment demand.

Conclusion

Predictive modeling has become central to business because accurate forecasts change everyday decisions. For an auto service chain, public data on driving, joined with the chain's own records, already show a forecasting method that halves the error of current practice, a strong case for the full model in the final project.

References

Federal Highway Administration. (2024). Traffic volume trends. U.S. Department of Transportation. https://www.fhwa.dot.gov/policyinformation/travel_monitoring/tvt.cfm

Loshin, D. (2012). Business intelligence: The savvy manager's guide (2nd ed.). Morgan Kaufmann.

Shmueli, G., & Koppius, O. R. (2011). Predictive analytics in information systems research. MIS Quarterly, 35(3), 553-572. https://doi.org/10.2307/23042796

U.S. Energy Information Administration. (2024). Weekly retail gasoline and diesel prices. https://www.eia.gov/petroleum/gasdiesel/