Call Center Scheduling

There are "of-the-shelf" tools for the scheduling of single-skill agents. The scheduling of multi-skill agents is theoretically possible using Monte Carlo simulation (see the section 2.7). However the simulation time is to large for an on-line scheduling. Thus, at the moment, the most convenient approximation to multi-skill scheduling seems to be some reduction of multi-skill problem to the single-skill one. One can do that by representing each multi-skill agent as a "weighted" single-skill one, namely: , and estimating the unknown weight by minimizing the squared deviation between the results of approximate single-skill model and the genuine multi-skill one.

(148)

Here denotes the results of the $k$ iteration of a Monte Carlo simulation of the multi-skill system, $\lambda$ is its call rate and is the number of multi-skill agents. By we denote the results of the $k$ iteration of a Monte Carlo simulation of a single-skill system where is the number of single-skill agents replacing the multi-skill ones.

The single-skill simulation described in section 2.13 can be generalized to the multi-skill case, too. In this case the greater computing time of the multi-skill simulation will be needed only for estimating the "optimal" weights for different call rates. It will not be an easy task but not the on-line one. The optimization methods included in the web-site (see section ). For example, the method was used to obtain the best values of .