Diversity of available tours ensures that each agent works five
days in each week, Saturdays and Sundays are adequately staffed,
and higher volume days employ more agents. In any feasible
schedule, some agents may be assigned to simple Monday through
Friday tours (from the set denoted R5 for standard tours or S5 for
split tours). Other agents will be assigned a Saturday or Sunday tour
(from the set R1 for standard or S1 for split), along with a weekday
tour with a nonscheduled (NS) day (from the set R4 for standard or
S4 for split). Each agent’s weekday start time will not vary within
the week. However, an agent assigned to an NS day tour may have a
completely different tour type and start time on the assigned
Saturday or Sunday.
Our modeling objective is to generate a tour distribution such
that a staffing level of at least ri is achieved on each interval, while
minimizing the sum of squared normalized deviations between
available staff and required staff. This objective recognizes the
‘‘diminishing returns’’ property of performance improvement as
staffing increases, and therefore distributes staff surpluses as
evenly as possible throughout the week. The ancillary restriction
on the frequency of split tours p must also be enforced, along with
permissibility of weekend split tours as controlled by binary
parameter q (1¼yes, 0¼no). To capture these requirements, we
formulate a quadratic program in which the decision variable xjk is
the number of agents assigned to tour j with start time kATj. Letting
auxiliary variable yi be the staffing level for interval i, we write the
formulation