Callback Strategy Optimization for Any Call Center
Callback technology is a fascinating topic. At one point in time if you wanted to do Callbacks you either programmed your own middleware or purchased a pricey third party product. These days, most contact center solutions include Callback functionality. Often there is little or no extra cost and the functionality can be turned on or off at will.
Back in late 1999 I had the opportunity to study the network level call details of Net Zero’s NBA advertising campaigns during the quarter time and half time intermissions of national play-off games. One of my recommendations at that time was to handle the spikes with call-backs. It worked great. Callbacks can be a useful tool to respond to brief intense advertising spikes.
Mis-Use of Callbacks
However, Callbacks are frequently mis-used and relied upon for all the wrong reasons.
Try calling the Air Canada call center. Callbacks are always-on and wait times are always over an hour. It’s a great example of using Callbacks instead of good decision making – with persistently terrible results.
Root Causes of Poor Callback Strategies
Much of the blame falls on the heavily-marketed, under-performing WFM solutions. These products mislead customers into thinking their forecasts are reliable. They are not reliable, not accurate, not even close. Technically speaking, they are not even forecasts – though that’s what everyone still calls them.
The high forecast accuracy reported by WFM solutions is entirely fake. That’s not a criticism of one product. It’s a factual statement about how every vendor manipulates data to present falsified metrics. Most planners don’t realize that this is happening — even though it’s been going on for 30 years.
Ilustrating Callback Strategy Errors
So let’s say you have the right number of staff (which most call centers do). But you have the grave mis-fortune of having those staff timed improperly (which many call centers do thanks to WFM).
As your day progresses, the queue gets larger and larger. Each interval has to deal not only with the calls that arrive, they also have to deal with the misfortune of every new call lining up behind lots of callers that did not get serviced in the prior intervals. The callers start to abandon and redial. When you look at your queues, you see large spikes in the wait times. But when you look at your WFM metrics, you see that the offered calls were right in line with what was forecasted. So it must be call spikes right?
Counting On the Wrong Numbers For Callback Stategy
Your WFM solution is only counting calls at the rate they are answered or abandon. So if you are Air Canada, callers are arriving at some rate that the WFM solution does not even look at. The callers wait for an hour. The WFM solution chooses to count the callers at the rate they exit the queue (as they are answered or abandon). That also happens to be the rate that the WFM solution prepared Air Canada to answer the calls.
Notice that calls can only be counted at the call answering rate. That’s why the forecast looks accurate. In reality, the forecast is so bad, it is the primary cause of the long wait time misfortunes.
Callbacks for the Wrong Reasons
But let’s say Air Canada never figured out that their WFM solution was faking the forecasts and faking the forecast accuracy.
If they assumed their forecasts were accurate, then on the surface it would make a lot of sense to buy a Callback solution.
After all, if the calls always come in at the forecasted rate, then the long wait times must be due to brief spikes??? Brief call spikes can be absorbed by a Callback solution. So what could go wrong???
Positive Environment for Callback Solutions
It may or may not be obvious that a Callback solution does nothing positive for a chronically understaffed call center. Callbacks are only useful if the demand rate spikes above staffing levels briefly, then drops below long enough for the call center to catch up.
That’s why Callbacks worked so well for Net Zero. A brief 3-minute spike around the commercial was followed by a long lull in calls. Agents flipped into Callback mode and did their best to empty the Callbacks before the next break in the game. It was great for maximizing both revenue and agent utilization.
The Callback Strategy From Thin Air
Call centers like Air Canada don’t use Callbacks in the proper context. They persistently understaff every interval of every day, and they rely on Callbacks in the misguided hope that things will get better. The only thing that gets better is the reported service level.
Every caller who accepts the Callback still waits the same amount of time. However the wait time happens outside of the queue. So service levels and wait time statistics are tainted to look better than they actually are.
Calls continue to be answered at the rate planned by the WFM solution. The WFM solution continues to produce forecasts around the call answering rate. That’s not really a forecast. Literally it’s not.
A real forecast should be constructed out of demand. When you take the average historical capacity (ignoring all signs of real demand) that’s not a forecast. It’s just a guarantee that the call center will never change or adapt.
So Air Canada is not our customer. If they were, their wait time problems would be solved pretty darn quickly and — they would use Callbacks in completely different way than they do today.
Callback Detail Analysis
When we take on a new customer, we promise huge improvements to wait times and productivity. Callbacks are never a component of reducing the wait times. If the customer happens to be using Callbacks this comes out in the discovery phase. If so, we take a very close look at how they are using Callbacks and we ensure that the detailed records of all calls (including Callback records) are included in the assesment data that gets processed. The results of the assesment is pinpoint becasue the second-by-second information pertaining to every call is intricatly analysed for its relvance to all questions of capacity.
Callbacks and the WFM Planning Cycle
Here is an interesting example of a customer that really needed quality capacity planning help. They used a Market leading WFM solution and their wait times kept getting longer and longer. It’s a vicious cycle. Because agent quit, training costs go up, the call center becomes less profitable and hiring budgets get cut. Several months before they dumped their WFM vendor they turned the Callback option on in their programmable ACD. I’m going to show you their Tuesday capacity over an eight-week period of using Callbacks. I picked Tuesday to take holiday Mondays out of the mix.
This is eight weeks of intraday on phone capacity. This has nothing to do with demand, it’s pure capacity. The demand is way way higher than capacity every day of every week. Notice that the on phone capacity gets weaker and more erratic from left to right.
It’s actually the Callbacks that are driving this. More and more resources get sucked out of inbound queues and into performing Callbacks. Callbacks are inherently less efficient because the calls take longer, the original caller is not always there, and the topic is not as fresh.
We asked this customer to turn their Callbacks off immediatley and to keep them off while they hire and train enough staff to satisfy their SCO forecast. It’s going to take them a few months to complete the hiring and training.
The SCO analysis of their call details made it very clear that Callbacks were harming the call center. So check out what happened to their on phone capacity during the next two weeks that the Callbacks were turned off.
Their on phone capacity exploded as soon as the Callbacks were turned off. There was no new hiring. In fact, they lost a few agents due to attrition. There is no tricky accounting going on here. We counted all of the agent conversations whether they were Callbacks or inbound calls.
When used improperly, Callbacks can really suck the life out of a call center.
The Hiring Question
The most typical situation that we encounter is that a call center with long wait times already has the right number of staff — its only the timing that needs to be corrected. This case study is about the less frequent situation. Some call centers have the unfortunate distinction of having fallen so far behind thier demand curves that substial hiring is needed.
To the Brink
Call centers that get improper guidance from workforce management systems suffer in multitudes of ways. That suffering accumulates over time and culminates in situations that can get pretty hopeless. When Workforce Management systems force call centers into distress mode, call centers are often tempted to turn to Callback solutions for help.
Risk Identification for Callbacks
It’s really important to know that when the forecast is persistently wrong, Callback solutions can quickly make things worse. It’s not the fault of the Callback solution. It’s a consequence of using spike smoothing technology in situations where the supply of agents is just never close to the real demand that unfolds across intervals.
WFM solutions never understand that demand because answered and abandoned call counts are ignorant of demand. They only understand how to perpetuate the average historical capacity.
Solid Answers About Wait Times and Callbacks
SCO technology can answer a lot of questions including how to effectively use Callbacks. Sometimes the most effective use is to turn them off.
SCO can immediately tell you how to time the staff so perfectly that even significant variations in demand are absorbed comfortably into a right sized queue.
What Not to Ignore
Right sizing a queue is something that no interval based WFM solution will ever understand because they assume the queue empties at the end of every interval.
Interval based forecasting has no concept of demand and no concept of queuing. Yet the very same math is marketed as highly sophisticate forecasting algorithms to support high forecast accuracy.
Planners have been indoctrinated into the belief that WFM solutions are good at making predictions.
Recognizing that none of it was ever true is the first step towards properly timing call center resources for consistently short wait times and record high productivity.
Making good decisions about Callback programs starts with recognising that the raw materials for the decision will never flow from software that relies on interval based data and falsified metrics.