Next-gen strategic crew planning

Optimizing the size of a railroad’s trainman and engineer workforce is vital. If there are too few crew members, the railroad risks failing to meet customer demand, resulting in customer service reliability problems and potentially translating into lost revenue. If there are too many crew members, the railroad risks incurring guaranteed payments, including the possibility of furloughs, as well as incurring avoidable training costs.

Understandably, many HR and crew planning executives protect against shortages, so they buffer. Hiring decisions are often informed by crew manpower models that lack the flexibility needed to support the business and the accuracy to drive increased efficiency. Creating a monthly forecast for many hiring locations can be a multi-step process that takes several days to complete.

At one Class I railroad, members of the business group who used the manpower forecast envisioned a next-generation system: a modern, data-driven application. They believed that correctly sizing the workforce will help better serve customers, manage assets and control costs—especially important attributes for railroads implementing PSR (precision scheduled railroading) practices.

Leveraging Data and Advanced Analytics

Princeton Consultants and the railroad’s business group partnered to develop best-practice processes, functional requirements and a forecasting methodology to lay the groundwork to build the next-generation strategic crew planning system. The project team interviewed subject matter experts to understand the crew hiring and training pipelines, and to identify key supply/demand factors. Historical data was analyzed to identify trends and select the most predictive model. For the demand-side forecast, the team opted for an approach based on crew starts that incorporated seasonality, based on the principle that crew demand is driven by a demand for starts (i.e., shifts worked) and the average efficiency of employees working the assignment.

The supply-side forecast considered the conductor and locomotive engineer training pipelines, less anticipated turnover. A factor analysis considered demographic variables potentially correlated with turnover, such as the employee’s age, number of years worked with the railroad, the employee’s gender and craft. A survival model was employed to forecast attrition and retirement at each hiring location, based on the composition of its workforce.

The Next-Gen System in Action

The resulting crew planning system informs strategic hiring decisions by predicting future staffing needs. It provides an intuitive user interface and simple workflows to accommodate business users of all backgrounds. A modern dashboard displays crew headcounts and other key results to facilitate decision-making and executive updates.

With a forecasting model at its core, the system recommends clear and actionable hiring targets. Forecasting is performed for each assignment to capture unique aspects such as historical reliance on the “extraboard” and other factors.

Business users may enter location-specific attributes, such as training durations and perceived risk, to customize the forecast. Risk is considered to appropriately size the buffer and the model’s recommended staffing levels. The training duration affects the time for employees in training to mark up, recognizing that some locations will produce qualified conductors faster than others. Capacity is also a consideration in that smaller locations can accommodate fewer trainees; as a result, it may take longer to realize the same workforce gains.

Users construct hypothetical training schedules and run what-if scenarios to predict the impact on future hiring needs. They design training plans today that will serve anticipated staffing needs 1-2 years in the future. They explore multiple scenarios in parallel, review them with stakeholders, and select the most strategic one to enact.

Users have considerable visibility into and control over input data. They configure look-back periods to target historical periods they believe are indicative of future activity. Outliers can be removed to cleanse data. Volatility is managed by smoothing natural fluctuations to achieve staffing levels that are realistic to implement. At locations where conditions are evolving, business users make ad hoc refinements of the model as needed. All changes and overrides are audited for future reference, promoting transparency and accountability.

Crew redistricting is a common practice in the rail industry. Typically, when hiring locations are consolidated, assignments are moved and re-named. Some assignments end and new ones emerge. To accommodate crew redistricting, business users manage assignments to preserve the integrity of the historical data which ensures a more accurate, reliable forecast.

Implementation and Getting Started

To encourage adoption, the team and project stakeholders engaged regularly with division superintendents and general managers, who often influence crew hiring decisions based on the perceived needs of their hiring locations. The new crew planning system was integrated into the existing decision-making process, so it would be viewed as a strategic planning tool rather than a prescriptive edict.

Based on the challenges and successes of this project, the following steps will help you transform crew planning and hiring:

• Identify which decisions could benefit from a data-driven approach, and whether sufficient data is available for analysis.
• Determine where human input has the potential to improve the forecast.
• Look for opportunities to streamline the decision-making process to be more transparent and accountable.
• To facilitate adoption, stakeholders should participate in the creation of an implementation plan to help integrate the new crew planning system into your hiring process.
• Emphasize data quality and integrity: invest significant time identifying reliable source data, and cleaning and preparing that data for use in the model.
• Validate results with stakeholders to sanity check values, building confidence in the forecast and facilitating adoption.

Early Results and Potential Gains

The business group executives use their new system to inform crew hiring decisions across their network. What used to be a multi-step, days-long process can now be accomplished in five minutes, the time required to run the model. The business now has visibility into the data behind the forecast and can explore what-if scenarios. Location-specific attributes let users manage risk and tailor recommendations to each hiring location.

To quantify potential benefits of transformed strategic crew planning at your railroad, consider this conservative scenario. Optimizing the size of the trainman/engineer workforce could allow a significant reduction in the extra-board buffer, perhaps 1% of the total workforce. If the workforce were reduced by 100 employees, savings in guaranteed payments and employee benefits would amount to more than $4 million annually. Furthermore, training costs would be cut by $4 million in the first year.

As this Class I railroad continues to roll out its next-gen system and measure the benefits, other railroads with similar goals and objectives should assess their crew planning and hiring process. Opportunities to contribute to an operating ratio improvement may be more attainable than previously thought.

About the Author

As a member of the Princeton Consultants Rail Practice, www.princeton.com/rail, Stefan Sabo builds data-driven software applications and provides systems integration solutions for Positive Train Control (PTC) projects. Stefan earned an MA in Mathematics from the University of Cambridge and an MA and BA in Mathematics from the University of Pennsylvania.