What needed to be understood.
Customer arrivals varied substantially by weekday, paydays, and holiday proximity. The analysis needed to separate these effects and convert coefficients into an understandable scheduling plan.
How the analysis was built.
- Created weekday dummy variables with Friday as the baseline.
- Modeled arrivals using staff payday, faculty payday, holiday proximity, and weekday indicators.
- Interpreted model fit, coefficients, and statistical significance.
- Translated forecast effects into high- and low-staffing days.
What the evidence showed.
- The model explained about 72% of observed arrival variation.
- Staff and faculty paydays each added roughly 369 arrivals.
- Holiday proximity added about 281 arrivals.
- Wednesday was approximately 479 arrivals below the Friday baseline.
What should happen next.
- Increase teller coverage on Fridays, paydays, and holiday-adjacent days.
- Reduce coverage on lower-volume weekdays while preserving service-level safeguards.
- Refresh the model regularly and compare predicted arrivals with actual wait-time outcomes.
What this project does—and does not—prove.
The analysis focused on arrivals rather than service duration, teller productivity, queue abandonment, or explicit service-level targets.