Ranvir: I don't agree with your analysis. We are discussing Occupancy vs. Service Level, not workflow. Let us examine your analysis. In Scenario 3 you mention "..calls are still coming and they are still getting idle agents..". This scenario, according to you, is created by a workflow analyst. How do you "control" the way random calls arrive in a Contact Center through a workflow? You can't! Your multi-skilling example, too, misses the crucial point. You are comparing idle time of one skill with the Service Level in another. You say idle time is more for agents in a skill that has no or less calls (say Skill A). The low service level is for another skill (say Skill B) which is getting back-to-back calls and consequently no idle time for agents. So your argument that Service Level is low for Skill B and occupancy is also low for Skill A agents, does not make sense because these metrics are not for the same skill. In fact, by your own token "more occupied" agents AND a lower Service Level is being experienced by Skill B, proving the inverse relationship between Occupancy and Service Level. Whereas if Skill A started getting more calls than what it is getting currently, then the Service Level would certainly drop, but the agents would become more occupied, again proving that for Skill A, too, the inverse relationship holds true. You should follow an Erlang C calculator, such as ICMI's QueueView Staffing Calculator, (as given in the link in my previous response) to do the "what-if" analysis for yourself. In fact your analysis is more applicable for a non-real time, non-abandoning contact type such as email, which does not usually follow the Erlang C formula. Through workflow you can "present" agents with another work item just as they are getting free from the earlier one. But then remember that emails do not have Service Levels, and so the "Occupancy v/s Service Level" discussion doesn't apply to emails. Look forward to more views on this. Thanks -- Rajas Daithankar, Avaya Australia

Broadly speaking, yes, occupancy is inversely proportional to service level. But there are cases when there is a rise in occupancy with rise in service level and vice versa. Please examine 3 scenarios: Scenario 1 - calls are coming in such a way (whether in a neat packet or at random) that they are easily getting an idle agent to answer these calls. It means that the service level is up and occupancy is low. Again, in scenario 2 - calls are coming in such a way (whether in a neat packet or at random) that they are not getting an agent to answer them. It means either they will be abandoned or they will be answered after threshold. Clearly, occupancy shall be high and there shall be a dip in service level. Now, in scenario 3 - this is the scenario that the calls are still coming and they are still getting idle agents to pick them up but the impact of such call flow is a decrease in idle time of the agents. This environment is that stage which is to be created by a work flow analyst -- here the service level is also high and due to decrease in idle time, occupancy is also high. All calculations, be it by work flow program, Erlang or manual are for achievement of this phase. Had occupancy always been inversely proportional to service level then there was no need of work flow analyst in the BPO sector. This is that position where the company controls its expenses and hence increase in profitability. In all manpower calculations we do have to consider the occupancy factor as to how much output we can get from an agent. Normally the occupancy taken to calculate the calculations for manpower is 80%-85%. And similarly schedules are prepared taking occupancy factor under consideration. Now, as stated above, all the workflow programs, Erlang, and manual calculations are for calculation of call forecast and hence planning of FTE (Full-Time Equivalents) for the interval, days, weeks, etc. One more case when the occupancy and service level are not inversely proportional is when you have multiple skill sets and the agents are not trained to answer calls of other skill sets. Here it can also happen that there is no or less calls in a particular skill set so more idle time for these agents and on the other hand one skill set is getting back-to-back calls and hence no idle time. Clearly in this situation, service level will be low but occupancy will also be low. So in new environment there is a need for those agents those are trained to handle multiple skills. -- Ranvir Wadhwa, Teletech (Bharti) India Ltd.

Hi: The answer is more intuitive than explanatory. It is like "knowing" that 2 + 2 = 4, but to explain it is more difficult. However, let me attempt. Assume that you are staffed to meet a certain Service Level - say 80/20. If you now wish to provide a higher Service Level (say 90/20 or 80/10), you require more resources. Most people believe (wrongly), that if they do not increase the "bodies in chairs" (a.k.a. FTEs) they should still meet this higher Service Level by having the same number of agents work harder; and consequently they should realize higher "occupancy" as well!! Well, that would have been true, had the calls arrived in neat packets, each new one arriving "just in time" as a busy agent was wishing goodbye to the existing contact. Unfortunately, that is the last thing the calls do - and against all your natural urges for them to do otherwise, they arrive randomly. This is where the Erlang C formula enters the picture, and cutting a long story short, it calculates for you the number of "bodies in chairs" required to meet the given Service Level for the given Call Volume for the given Average Handling Time. (And you can use ICMI's QueueView Staffing Calculator or equivalent to see this for yourself.) In essence what it is telling you is this - for the same call volume, if I require to meet a higher Service Level, then I will require more resources to do so; AND in the process those resources will be 'more available' (than a lower number of resources handling the same call volume but for a lower Service Level) to handle the probability of the next random call. As you know intuitively, 'more available' means that they will be less occupied. Hence, the inverse relationship between Occupancy and Service Level. You might refer to my response to the question "Convincing an Erlang Non-Believer" under the Scheduling section of Queue Tips, for a related discussion on Erlang formula. Any comments are welcome. Thanks. -- Rajas Daithankar, Avaya GlobalConnect Limited