Cisco Cisco IP Contact Center Release 4.6.2 Leaflet

To calculate this factor, divide the number of agents required from Erlang-C by the productive agent 
percentage (or 1 minus the shrinkage percentage). For example, if 100 agents are required from Erlang-C 
and the shrinkage is 25%, then 100/.75 yields a staffing requirement of 134 agents.

Consider the following design factors when sizing call center resources:

Compute resources required for the various busy intervals (busy hours), such as seasonal busy hours 
and average daily busy hour. Many businesses compute the average of the 10 busiest hours of the 
year (excluding seasonal busy hours) as the busy-hour staffing. Retail business call centers will add 
temporary staff based on seasonal demands such as holiday seasons. Run multiple interval 
calculations to understand daily staff requirements. Every business has a different call load 
throughout the day or the week, and agents must be staffed accordingly (using different shifts or 
staffing levels). Customer Relationship Management (CRM) and historical reporting data help to 
fine-tune your provisioning computations to maintain or improve service levels.

When sizing IVR ports and PSTN trunks, it is better to over-provision than to under-provision. The 
cost of trimming excess capacity (disconnecting PSTN lines) is much cheaper than lost revenue, bad 
service, or legal risks. Some governmental agencies are required to meet minimum service levels, 
and outsourced call centers might have to meet specific service level agreements.

If the call center receives different incoming call loads on multiple trunk groups, additional trunks 
would be required to carry the same load using one large trunk group. You can use the Erlang-B 
calculator to size the number of trunks required, following the same methodology as in the 

. Sizing of required trunks must be done for each type of trunk group.

Consider marketing campaigns that have commercials asking people to call now, which can cause 
call loads to peak during a short period of time. The Erlang traffic models are not designed for such 
short peaks (bunched-up calls); however, a good approximation would be to use a shorter busy 
interval, such as 15 minutes instead of 60 minutes, and to input the expected call load during the 
busiest 15 minutes to compute required agents and resources. Using our 

, a load of 2000 calls in 60 minutes (busy interval) requires 90 agents and 103 trunks. We 

would get exactly the same results if we used an interval of 15 minutes with 500 calls (¼ of the call 
load). However, if 600 of the calls arrive during a 15-minute interval and the balance of the calls 
(1400) arrive during the rest of the hour, then 106 agents and 123 trunks would be required instead 
to answer all 600 calls within the same service level goal. In a sales call center, the potential to 
capture additional sales and revenue could justify the cost of the additional agents, especially if the 
marketing campaign commercials are staggered throughout the hour, the day, and the various time 
zones.

Consider agent absenteeism, which can cause service levels to go down, thus requiring additional 
trunks and Unified IP IVR queuing ports because more calls will be waiting in queue longer and 
fewer calls will be answered immediately.

Adjust agent staffing based on the agent shrinkage factor (adherence to schedules and staffing 
factors, as explained in 

Allow for growth, unforeseen events, and load fluctuations. Increase trunk and IVR capacity to 
accommodate the impact of these events (real life) compared to Erlang model assumptions. 
(Assumptions might not match reality.) If the required input is not available, make assumptions for 
the missing input, run three scenarios (low, medium, and high), and choose the best output result 
based on risk tolerance and impact to the business (sales, support, internal help desk, industry, 
business environment, and so forth). Some trade industries publish call center metrics and statistics,