Originally published in Contact Center Pipeline, September 2017.
Skills-based Routing (SBR) is a feature found on any ACD today, and it’s been around for over 30 years. It is so common, in fact, that it’s become an afterthought when purchasing a contact center phone system: Auto attendant? Check. On-hold messaging? Check. Skills-based routing? Check.
Hold on a second. That check mark next to SBR makes it sound like a simple, low-value, one-size-fits-all feature that requires little in terms of expertise to deploy. Nothing could be further from the truth. SBR comes in many forms, is exceedingly powerful, and may be the least understood and most error-prone piece of technology you can find in a contact center (for a classic example of how to make a huge mistake with SBR, read the example in call-out 1). Understanding the various forms of SBR is critical to deploying it in your environment.
Forms of SBR
While SBR can be approached in many different ways, the easiest (and probably most important) distinction to make is between true SBR and multi-skilling. While often passed off by the vendor as SBR, multi-skilling is really just the ability to allow an agent to take calls from two or more different queues. Think of the classic center that has both a sales line and a service line. No one wants to abandon sales calls, so the more experienced service people might be given the sales skill so they can receive sales calls along with their service calls. While that’s a nice feature, if they are undistinguished from the sales-only team, the potential value is limited.
A more advanced approach, utilizing true skills-based routing, recognizes that these dual-skilled agents are a second priority for sales calls. This is handled through some form of a skill rating that allows the system to check for an available sales-only agent first before considering the dual-skilled rep.
While the skill ratings allow for more flexibility than simple multi-skilling, it comes with a limitation that often goes unmentioned by the phone system vendors: skill ratings only work in an agent surplus condition. Once calls queue up (a call surplus condition), the ratings no longer matter, since the next rep that becomes available with that skill will get the call, regardless of how they are rated. For SBR to work under both agent and call surplus conditions, the skill ratings have to be augmented by some more advanced programming capabilities that may or may not be presented as SBR. In our sales-service scenario, examples include:
A delay that holds the sales call in queue for x number of seconds when a dual-skilled agent becomes available, in the hopes that a sales-only agent becomes available during that time
A “reserve” option that only makes the dual-skilled agent available for the sales call only if there are, for example, 2 or more service agents available (effectively “reserving” them for service calls)
The SBR Balancing Act
So let’s assume you have the full complement of SBR options that can get the call to the best possible agent under either call or agent surplus conditions. What you have, then, is one very powerful tool with nearly limitless possibilities. And therein lies the problem. The more intricately you design your SBR, the more impossible it becomes to predict what will happen.
Harnessing this power requires two key ingredients. The first is expertise. The vendor may be able to supply this, but most vendor resources dabble in the shallow end of the SBR pool, given that they do not know your business and have to make sure they avoid call flow design disasters. More than likely, the expertise you need will come from your staff, since they better understand the impact to the organization in areas such as:
Incentive program implications
Value of reduced training efforts
Impact of occupancy rates that vary by agent
The second key ingredient is patience. Mistakes occur when too many programming options are implemented (or revised) at one time. Start small, addressing only the agent surplus condition. Measure results, and adjust to find the right combination for you. Then start working on options for the call surplus condition. Take a conservative approach, and again measure at regular intervals to make sure you are getting the desired results. The types of things that need measuring include:
Occupancy rates by agent – SBR changes this, and while it is good to keep your best people occupied, you don’t want it to get too imbalanced
Service level by call type – If you only look at combined numbers, you might miss a caller segment that is being penalized by the SBR design
Quality rates – Used too aggressively, SBR can have too many calls handled by agents that are a second, third or fourth option
Keeping all results in balance is the key, and to do that you need to be vigilant. Seemingly unrelated events, like adding a large training class of single skilled agents to the mix, can affect results without ever touching SBR parameters.
Let’s face it – a deep, intricate understanding of SBR is in short supply these days. Contact center leaders shouldn’t be expected to possess this knowledge, but they do need to understand just how important this feature is and how the right balance is so important to the organization. If the expertise exists on your team already, make sure you value it. And if it doesn’t, spare no expense to get it. The payback is well worth it.
Call-out 1: The heavy price of SBR errors
Here’s the scenario: 20% of an organization’s callers are Spanish-speaking, while the remainder speak English. Likewise, 20% of the staff is bilingual, with the rest speaking only English. The company’s ACD offers SBR, so the routing is a no-brainer, right? Spanish calls go to the bilingual team, while English calls go to the English-only group. Of course, you want to keep the queues balanced, so if no one in the target group is available, the call will overflow to the other team (English to bi-lingual, Spanish to English-only). When an Spanish caller overflows to the English group, just dial up a translation line. Yes, that’s expensive and the call takes twice as long, but it won’t happen all that often.
Or will it? That all depends on the percent of time a queue is present. In this organization’s case, staffing situations caused a nearly perpetual queue, from open to close. What happened? When a bilingual team member became available, there was an 80% chance that the next call was English speaking. That eventually led to 80% of the bilingual staff handling English calls, while 80% of the Spanish callers were answered by English-only agents. Translation costs quickly rose to tens of thousands of dollars a month, and AHT went up over 15% - which of course led to longer queues that perpetuated the cycle. A classic example of an SBR Fail.