Agents in a contact center are grouped dynamically by defining performance metrics against which agents are to be judged, and generating an evaluation score for each of a number of agents by comparing their actual performance data to the defined performance metrics. Agents are assigned to the groups based on their performance scores. This provides the ability to generate reports and route contacts according to actual agent performance judged relative to user-specified performance metrics. The sizes of groups can be adjusted in real time to take account of actual or anticipated levels of contacts with different requirements, and agents can be reallocated among the resized groups on the fly based on the ranking of their evaluation scores.