Amanda Madison

Hi! 
Would this mean that if for example an agent does 6 chats (engagement time for each is 15 min) in the first 45 minutes and then in the last 15 minutes of the hour goes on a break, the concurrency would be 6x15/60 = 1.5 chats instead of 2 chats only because they weren't active the whole hour? Or does it limit to time agents are “online” and ready to take new chats? 

Tymeshift staffing forecast is based on Erlang C model. It effectively determines staffing needs for chats or phones with quick response times, but it falls short for deferred transactions with long cycle times. For instance, adjusting the FRT targets in Tymeshift's forecasting system from 1 hour to 24 hours doesn't change much, though it should (even if we have selected e-mail as the channel in staffing parameter section). It will always suggest that more agents/hours are required than in reality. 

Due to this issue, we can't rely on Tymeshift's forecasting system to accurately predict the required hours for deferred transactions. And we must manually schedule email time, which is a time-consuming process.

Our suggestion is to incorporate Workload model specifically for these deferred transactions. The Workload model is the simplest methodology for calculating staffing or schedule requirements. It multiplies the forecasted volume by the forecasted transaction time, for the chosen time period and divides it by time. Other factors as occupancy, SLA and shrinkage can be incorporated in the formula. 

To summarise, the Erlang C model is less effective for long-cycle transactions like emails because it assumes immediate service initiation, relies on constant queue dynamics, and works best with consistent service times. Long-cycle tasks, however, have irregular arrival patterns, varying priorities, and flexible response times, making accurate staffing predictions challenging. To boost up the staffing accuracy in all channels we require diversity in the models we use to predict required hours in different channels.