Skill Based Routing Challenges

[JPalkowski]

We are finding it necessary to look at skill-based routing to address some of the complexities of our inbound customer care center.

I would be interesting in hearing what some of the challenges have been in implementing skill-based routing in your call centers.

Our biggest concern is scheduling. We run a 24 X 7 center, and our concern is forecasting and scheduling the right number of agents with the right skillsets. We would likely have about 4-5 different skill levels.

If anyone would like to share their experiences on how you handled this challenge with your call center, please reply.

Thank you.

[Mark Brannan]

We have been using skills based routing for a few years - we are a technical software support organization so the comparison may not be apples to apples.

Obviously the biggest challenge is what you stated, how do you plan, staff and forecast to accurately achieve service level targets with many skills. The only answer I can give you is via a workforce management software tool. We use Blue Pumpkin and are very happy with the results. They also have a planning tool which allows me to run sample scenarios at the click of a button to see how to further maximize efficiencies. This may or may not be the right tool for you - it just works for me.

Besides that, the challenges relate to training, continued education, and eventually end up with a discussion about how many skills are too many for your agent to handle. If you are going to have 4-5, it may be reasonable for you to expect agents to pick up all of these skills and remain very efficient at solving all of the problems. We are beginning to hit a max on skills though, where people actually become less productive by picking up another skill, since they can't solve the questions as quickly as they once could - too general at this point.

Let me know if you think it is worthwhile to continue this discussion.

Mark

[olsonbr]

Currently in one of our 24X7 environments we have a client that has 5 skills of traffic that are handled by multi-skilled agents. We forecast our calls on the indivuduals lines and then consolidate them to determine staffing needs.
One factor that has a major impact on call center staffing is the size of the center or the agent group. Centers handling larger volumes of calls will naturally be more efficient than smaller ones. This is due to the economies of scale of large groups.

As seen in the example below, doubling the call volume does not require two times the number of staff to meet the same service goal of 80% in 20 seconds. And when call volume increases eight-fold, only about six times the number of staff are needed. As the volume grows, the staff to workload ratio gets smaller and smaller.

Calls per Hour Workload Hours Staff Required Staff Workload Ratio Staff Occupancy
(workload/staff)

100 8.33 12 1.44 .69
200 16.67 21 1.26 .79
400 33.33 39 1.17 .85
800 66.67 74 1.10 .90
1600 133.33 142 1.06 .94


The reason for these increased efficiencies and the lower staff to workload ratio is simply that with a higher volume of calls, there’s a greater likelihood that when an agent is finished with a call, there’s another one coming in right behind it for that person to handle. With a bigger volume, each person has the opportunity to process more calls each hour. Each person spends less time in available state waiting on a call to arrive – in other words, they’re busier or more occupied with call workload.

[JoeB]

Staff-to-call-volume ratio depends on multiple factors. It is, as stated above nonlinear, although the reason is not quite as stated. Rather, calls arrive randomly in a statistical distribution that is unrelated to the rate in which call are dismissed (resolved). The statistical model is quite involved and rules of thumb donÂ’t work. See http://www.diagnosticstrategies.com/...c_modeling.htm for a bit of theoryÂ…

As for efficiencies, the more separate queues are involved, as is often the case in skill-based routing, the less efficient the entire operation is. This has little to do with the size of each group, but rather the fact that the agent utilization, by definition, is less than 100%, and the more queues there are the more internal inefficiencies exist. If calls cannot flow to an available agent (because s/he is in a different skill-queue), there is less probability to take advantage of available resources (which may also decrease service levels).

So put simplistically, skill-based routing can deliver better quality, but at the cost of increased headcount and decreased efficiencies. The challenge, as stated above, is to find an optimum.

Workforce management software and similar forecasting/planning tools are used to create what-if scenarios to estimate resources and estimate the service levels they can provide. There are numerous such tools with varied capabilities (and prices). You can try http://www.diagnosticstrategies.com/easyerlang.htm to understand the functionality of such tools and select the one that works best for you.

[JPalkowski]

Thanks for the replies. Yes, I find this discussion to be very worthwhile.

I am curious, what mechanism are you using to route calls by skills? Do you have different incoming numbers, do you rely on callers to select the correct prompt on a voice menu/voice response unit, or are you using some other method?

Our initial thought is to use our phone switch with a basic voice menu to route calls. We are concerned that many customers may not choose correctly, and we'll have the issue of having to transfer calls between agent groups.

I would be interesting in understanding how you and other centers have addressed this.

[Mark Brannan]

Jeff, we use the switch and phone menus and allow our customers to route themselves. Easily over 90% of our customers select the correct queue - so internal transfer is really not an issue.

I agree with the assessments above - in particular that skills routing increases quality while reducing efficiency. I am constantly looking to determine if I can "combine" separate lower volume queues for this exact reason.

It also becomes an easy answer for development and even some in our own department on how to support a new application - simply make another queue. But I won't do that since it is so inefficient - and that brings up another problem with skills, the small volume queue (although technically you end up with concerns on the other side of the coin if you have a low volume application and need to train your entire staff on it).

Mark

[mdbabb]

We also use skill based routing within our call center and utilize multiple methods for this. Our call center is an HMO, and thus we have slightly different methods than the traditionall call center.

In our clinical areas, we utilize agents to triage member issues using medically staff defined protocol to route to specific skill groups. These agents that receive these calls are cross trained in other skill groups in order to increase efficiency yet still provide increased quality of care. We do this through the use of an Avaya (Lucent) Advocate Software, that allows us to assign skills and priorities to our agents. For example, an experienced OB nurse would take primarily OB calls, but also take internal medicine calls when OB call volume is down. Most ACD systems offer this capability though sometimes this is considered an additional pay feature.

We also have a Member Services department through which callers utilize an automated attendant tree to select the area for which they are calling. We have had a high success rate with this as well and found that the tree options allow us to better forecast and trend caller and staffing needs for each area. This is extremely helpful when issues arise that require temporary staff that are trained only on the specific area to address in order to handle the temporarily higher call volume.

We are currently not using skill based forecasting in our WFM system, but will be implementing this in the near future. Due to the nature of the cross training, we are still able to forecast FTE needs.

I also have experience in implementing a segmented skills based routing function for a telecommunications carrier. That company implemented different service levels for different levels of companies (e.g, the largest customers or new customers, depening on sales needs) where identified as VIP customers through the CRM system, which interacted with the IVR and routed calls to the approriate skill. Each skill had different service level requirements (e.g, Platinum customers hit a skill with a service level of 90/10 while standard customers would be routed to a skill with a 70/60 service level.) This was challenging to implement due to back end system issues in the legacy billing systems where parent child relationships over multiple billing platforms were not clearly defined.

In larger call center environments (200 seats+), skill based routing can offer callers increased quality and still maintain high efficiency and occupancy environments. However, in smaller centers (under 100 agents), it can be difficult to realize to the increased quality and still maintain good occupancy and agent utilization. Skill based routing in small centers can be almost impossible to attain without historical data (either through a CRM or reason code system) and accurate Work Force Management systems. We utilize IEX Totalview in my current position, and TCS at my previous company. Both systems offer excellent tools for forecasting, staffing, and tracking of skill based routing systems.

[npop]

Has anyone every had to staff a call center with Customers with various service levels i.e., 80% :20; 90% :30; :60. Is it best to use a combination of skill based routing and pooled - or would grouping by service level be the best way to keep headcount lower? Seems to me that staffing by service level would increase the headcount needed to answer the calls and meet all of the service levels.

[JoeB]

This is a very interesting question, which cannot be answered completely without additional, site-specific data.

Assuming that the all the calls are identical, i.e. the same staff can answer calls independent of the service level requirement, then staffing for the highest service level, considering the call volume of the entire population will yield a lower head count than if you create a separate groups for each service level and route calls accordingly. In the latter case, each group will be better optimized for its own call volume, but overall the call center will operate in a lower utilization. In other words, it may be more cost effective to allow customers to enjoy a higher level of service than they have actually signed up for (!) but this has to be determine by looking at the call volumes from each group, associated costs, penalties for not meeting the SLA, and how much extra capacity is built into the proposed resource allocation model

However, the above assumes that by “skill” we mean “proficiency” or “level of service”. If indeed different technical skills are required, then in many cases the inefficiencies created by the different queues are less easily avoidable.