Solving a Bank Staffing Problem Results
in 20% Staff Reduction

by Steve Mueller, Project Manager

DPA was engaged by Massachusetts-based bank to develop a staffing and scheduling system for tellers and customer service representatives (CSR) for each of their branches. The objective was to reduce their operating costs while maintaining service levels. Models for this type of environment (similar to call centers or ticket counters, etc.) require, at a minimum, a key volume indicator of work (KVI), a customer arrival profile, a service level threshold, and work-to-time standards.

Transactions were chosen to be the KVI for teller activity. There was no differentiation made between type of transaction because (1) there were too many to be practical in a rolled up ‘equivalent’ transaction, (2) not all transactions of a given type were equal (a paycheck deposit is not the same as a commercial deposit), and (3) the bank’s information system could not satisfactorily differentiate the transaction mix on a frequency needed for useful management reporting.

CSR activity was to be predicted by New Accounts volume, which is the primary measurement for this position.

To obtain the customer arrival profile, we asked the tellers to maintain a customer count by hour of the day for a period of two months (there was no customer count information in the bank’s system, unlike an ACD (Automated Call Distribution System) in a call center). We then established an average customer count by hour and day of the week. The tellers also provided a count of other, back room activities which were routinely conducted on a daily basis (e.g., ATM envelopes, night bags, etc). CSRs also provided a picture of customer activity by logging the number of customers requiring their services on each of twenty-one tasks over a period of one month.

To address the bank’s service goal of 100% of the customers served within 5 minutes, we chose not to employ an algorithm such as Erlang to establish staffing (customer volume and teller numbers are very small compared to call centers or airline ticket counters and Erlang tends to drive staffing high in low volume scenarios). Further, the bank was willing to accept the occasional service breach if there were savings to be had.

We developed a simple Wait Time Matrix for the teller supervisors to use to compare customers in line with open teller windows to see if any of the customers were likely to wait 5 minutes or more. If possible, the supervisor will redeploy people from back room functions to open windows to cope with peak situations. In this way the supervisor can take steps to meet the bank’s service goal.

Work-to-time relationships were developed through direct observation of customers at teller and drive-up windows and reviewed with bank management, all of whom worked as tellers in the past. Customer Service Representatives provided their own estimates of task time requirements since direct observation would have been difficult to do effectively without intruding on the privacy of the customer.

The teller staffing/scheduling model first calculates the ratio of average daily transactions to customers and other back room activities. This serves as the planning factor that can be used to convert forecast transaction volumes into customers and other activities. Then the model spreads these customer and activity volumes across the hours of each day based on the customer arrival profile. These hourly volumes are then converted into required hours by multiplying them by the established work-to-time standard hours per anticipated customer and other activities. The hours are rounded up and expressed as a number of tellers required each hour to handle windows, drive-up and other activities. The model also has flexibility built in to predict requirements for forecast peak days as well as peak hours on any given day.

The CSR model also establishes a baseline set of ratios between the KVI and each of their historical activity volumes. This ratio is used to predict activity volume for a given KVI volume. Since the CSR position is salaried and staffed by professionals, the bank was interested in understanding their capacity in this position so they might best allocate their resources between the branches to provide the right level of customer service without incurring unnecessary cost.

The teller model was successfully tested at 3 branches for several weeks. The branches participated fully and after some initial fine-tuning, the branches were able to achieve their goals and make a 20% reduction in tellers’ staffs (who were used to fill other open positions in the bank). The model is being roll out across the entire branch network.

DPA . . .   Articles  |  Case Studies  |  News Releases