Articles

No Tricks

SCO should be really easy to understand.  Better, more detailed information naturally produces much better results than incomplete and misleading information.  Yet, the most experienced people in the call center industry have great difficulty believing anything could produce a better plan than the established methods.

Interestingly, people from outside the call center industry seem to comprehend SCO with great ease.  So I invite you to take a quick vacation from what you do and look at SCO from a different perspective.  At first, you might think this story is about a card game. It’s not.

While vacationing this year, my wife and I met a couple from California named Christine & Whitney.  Late one evening the couple was playing cards on the main balcony.  They invited me to play but I had to admit that I only knew one game called “Cheat”.

As I explained the rules Whitney realized the same game had a different name in the US.  What Canadians might know as “Cheat”, Americans are more likely to know as “Bull****”.

The core strategy of the game is to be untruthful without being caught.  Each player pulls cards from their hand and places them face down.  The player states what the cards are and attempts to empty their hand without getting called on cheating. If you get caught, you pick up all the cards that the other players have discarded.  If you call cheat on a truthful play, you pick-up the cards.

With the discard tray more than half full, I played one Jack, face down and called it a Jack.  Christine called cheat and picked up roughly 25 cards.  I thought she was bound to lose but for the rest of the game she called cheat with pinpoint accuracy.  She had filled her hand with accurate information that gave her control of the game.  Her opponents soon realized that they had little choice but to play truthfully or pick-up. Christine won.

So with the game complete the conversation turned to our professions.  Christine is a mergers and acquisitions manager in the Automotive industry.  Whitney is a process engineer in the mass transit sector.  I explained that my company was involved in forecasting and scheduling for call centers.

Whitney said “oh, so you forecast call volume”.  I said “Not really, everyone else in the industry forecasts call volume, we forecast how many agents you need to satisfy the demand properly”.  Whitney asked if there was a difference.

So I explained that for the past 3 decades, the WFM industry had been playing a game that was a lot like “Cheat” except that very few customers had ever thought to look at the cards. The industry forecasts how much call volume to expect.  When the actual call counts come in, the industry makes-up numbers that match the forecast and tells customers that the forecasts were accurate. Customers never think to call cheat so they never see what’s under the cards that have been placed face down.

The couple was surprised, skeptical but interested. They asked how it would be possible for an entire industry to do this without being caught…after all call volume is call volume! Surely it’s not possible to make up your own numbers.

I replied, “Actually, it is easy, all they had to do was to convince customers to count calls the way that they wanted the calls counted.  You see all the experts in the industry and all the companies that sell this type of software tell customers that you must forecast with answered plus abandoned calls.”  Then I explained the basics of Offered Calls, Answered Calls and Abandoned Calls.

Whitney’s eyes were racing back and forth.  I could tell he was a problem solver and that his brain was working hard to process the contradictions.  He paused and then said “But how does that make it possible to change the numbers so a forecast looks accurate.”

I said, “Here, let’s play a game.  I’ll predict the number of calls that get offered to your call center between 9:00 and 9:15.  Then I’ll prove to you that my prediction was 98% accurate. You get to choose how many calls actually come in.  The only thing you have to do is guess high, because my prediction will be low.  Are you ready?”

Whitney said “sure, go ahead”.  I said, “I predict 300 calls and I’m going to tell you to put enough staff in the call center to answer only 300 calls.  Now, how many calls actually came in?”  Whitney said “400”.  I said “Perfect.  So you answered about 300 calls because that’s the maximum your staff could answer.  Roughly 2% may have abandoned, that’s 6 calls.  The Answered plus Abandoned totals is 306 so…TADAH!  My prediction of 300 calls was 98% accurate.”

Christine said, “That’s not true, Whitney told you there were 400 calls.  That’s 33% more than you predicted.”

I said “You are correct.  There were 400 offered calls.  I prepared you to answer only 300 of them so I only counted the 300 that you answered plus the very small number that abandoned.”

Christine asked me to explain what happened to the other calls.  I said, “No problem, they get counted in the next interval — at the rate that I have prepared you to answer them, plus the few who abandon.  Every prediction that I make will be accurate to within the abandon rate.”

So Whitney said “WHOW, so it really is like cheat”.  I said, “It really is like cheat”.

Whitney said “but the extra 100 callers are going to wait a really long time to get serviced.  They will form a huge queue and that will increase the wait times for everyone who arrives in the next interval, maybe even for the rest of the day.”

I said “You are right.  But that does not affect the accuracy of my prediction, its still 98% accurate.

Whitney shook his head and said “So the wait times don’t matter”.

I replied, you are correct, the wait times don’t matter.   They don’t matter to the forecast accuracy and they have no bearing on the next forecast. Each forecast is calculated exclusively using call counts and talk times. Wait time is not an input to the forecast.

“Well that sucks” said Whitney.

Yes it does” I said.  “Its much worse if your business is growing.  The more your business tries to grow, the more the wait times will increase. But answered plus abandoned call counts will keep feeding you the same forecast that always looks accurate.

Some customers will wait a really long time but that won’t show up in a service level of 80% in 120 seconds because the longest 20% of wait times don’t count towards the service levels.  And it likely won’t show up in your average wait times either because if you service some customers instantly and other customers wait a really long time, the average will look ok.

Luckily, the customers who wait the longest are very likely to switch to another provider.  That keeps the system in balance.  As long as your customers get fed up with you as fast as you can acquire new customers, those answered plus abandoned call counts can keep giving you the exact same forecast and it will continue to look accurate.

Whitney said “So if customers don’t leave, the abandon rate goes up and forecast accuracy drops.”

“Well..” I said, “That might be true if the industry actually counted all of the abandoned calls but they don’t.  They only count some of the abandons.”

“What, how can they get away with that?” asked Christine.

“So….”  I continued, “The industry has created different categories of abandons and they tell customers not to count some categories. If a caller hangs up too quickly, they call that a short abandon. Short abandons don’t get counted at all.  Typically the short abandon threshold starts off at 5 to 10 seconds.  However, when the total abandon rate increases, customers are advised to increase the threshold to 30 seconds or higher.”

“So the worse it gets, the more data gets discarded in order to boost the forecast accuracy.”  Said Whitney.  “Yes.”  I replied.

Whitney said, “that’s going to ruin a call center’s business.”

I replied, “Well it will certainly keep it from growing. And that’s the problem with forecasting with answered plus abandoned calls.  It makes the forecasts look accurate because it ties all call counting and all future forecasts to you current capacity.

Whitney shook his head in astonishment and laced his fingers behind his neck.  “So how do you do it then?  I guess you forecast with the real call counts?  The 400 calls that were actually offered?”

“Absolutely not” I replied.  “That’s even worse.  Let’s say 50 of those 400 calls arrive very close to 9:15.  If those are 10 minute long calls then that’s 500 minutes of work that need to get done substantially in the next interval.  Instead it gets counted in the interval it arrives.  That would make you grossly overstaff for 9:00 and grossly under-staff for 9:15.

True Offered Call Counts produce ridiculous looking forecasts.  If you tried to schedule to those forecasts it really would ruin your business.  One interval you would have dozens of idle agents waiting for calls that actually need to be processed in the next interval.  But in the next interval you would have staffed assuming the 500 minutes of work did not exist.  You see its not that the industry wants to cheat, they don’t have any choice. Answered plus abandoned call counts ties a customer to their current capacity which is less harmful than asking the customers to schedule to a ridiculous and unstable True Offered Call Forecast.”

“I see” said Whitney. “The call counts have a built in error rate and the error rate is huge.  If you have five minute calls, it’s a 30% error rate.  If its 10 minute calls then the error rate is 60%.  And the error rate is not random.  If the error is positive 60% for 9:00 then the error rate for 9:15 will be pulled towards negative 60%. The errors would toggle like that throughout the day. The staffing levels implied by a true offered call forecast would fluctuate between way too high and way too low.  It makes perfect sense.  So how do you avoid those problems?”

“Well” I said, “We don’t have any of those challenges because our software does not put any faith in any type of call count.  Instead, we have an algorithm that analyses the second-by-second real activities in any call center.  The software intricately understands the details of what is causing the long wait times and how to schedule staff to absorb the exact flow of calls across intervals. No matter how bad the wait times are in any call center, we can usually fix it overnight. Typically the call center does not need any extra agents, they only need a better forecast.

In fact, often they have more than enough agents.  When you service customers promptly, the conversations tend to be more productive. Short calls and fewer repeat conversations means less agents can satisfy more customers.  We’ve seen wait times drop by 64% and that tends to translate into a 20% increase in agent productivity.

Whitney said of course.  “You avoid the two minutes of the customer complaining about how long they waited so agents get more work done.”  I congratulated Whitney for his observation “Yes, you are exactly right. However it’s not just the first two minutes of complaining that is unproductive; the entire conversation tends to be underproductive.  Unhappy callers spend less and are less cooperative towards solving problems.”

Whitney added “So whether the customer is calling to make a purchase or to solve a problem, the success rate ends up being much lower. Agent time is consumed and the end result may be a lost sale or another call to resolve the same issue.”

“Exactly right”, I stated, “But when you remove the stress of those long wait times, productivity quickly returns to its peak.”

At that moment, our conversation was interrupted by karaoke night.  But that’s another story.

Written by

The author didnt add any Information to his profile yet

Leave a Comment