Frequently Asked Questions
Most planners who have been doing interval forecasting for a long time really want to believe that it’s been working for them. That’s perfectly understandable. It’s ok, we’ll walk you through all your questions so you can reconcile past and future.
Question 1 - Surely Interval forecasts do consider wait times. I always input a target service level or target ASA. Isn’t that a wait time input?
Actually, its not. The target is just a target. There is no historical foundation for that target when the only real data you input are interval call counts and average talk times. Also, keep in mind, those call counts aren’t real either, they are shifted to trace your historical call answering rate (via answered + Abandoned call counts). And true offered counts won’t work at all which is why no WFM vendor will even let you import them.
Question 2 - When you talk about answered plus abandoned call counts, I don’t know what you are talking about. I checked my ACD manual and Calls offered is not calculated as the sum of answered plus abandoned. It really is the number of calls offered to the Queue.
Yes, but what does your WFM solution import from your ACD. Wait for it…
Answered Calls and Abandoned Calls.
It’s been like that for decades.
So as long as you have been relying on interval forecasts, they have been tracing your call answer rate. That means they have been ignoring all evidence of change or growth.
Question 3 - I know of one WFM solution that does let you choose what type of call counts to use. One option is answered and abandoned counts. The other option is true offered counts as calculated by their ACD. That means you overlooked something.
There is one vendor that lets you import offered call counts from their own ACD. However, their ACD counts offered calls as the calls end. Another way to say this is “The offered calls are pegged to their end times”. So even though they call it offered calls, it actually the count of calls ending. So what you are seeing is the total of calls that are handled or abandoned each interval. Its marginally different from answered plus abandoned but not in any meaningful or helpful way. No matter how many calls are offered to the queue, “offered calls pegged to end times” will trace your call completion rate. It’s almost the same as your call answer rate because almost every time an agent completes one call, they answer another.
Question 4 - Well some WFM solutions do consider wait times because they have you input a wait time patience factor. How do you call that ignoring wait times?
Patience factors are a component of “Loss Models”. Often its just called “Patience” but sometimes its called “Patience and Loss Model” or “PALM”. The basic idea of a Loss model is to reduce call counts below the historical offered call counts. The lower the counts, the less staff you schedule. Some call that saving money. We call it strangling your call center and any revenue that it supports. The entire concept of Loss is doomed to produce longer and longer wait times. Pick any interval that has a high abandon rate. If 100 calls are offered and 10 abandon, the Loss model says why schedule staff to answer calls that are going to abandon anyway. So they reduce the 100 historical call counts to somewhere closer to 90. Staffing levels drop. Obviously, callers are going to wait longer and more will abandon. The sinister thing is that shifty call counts will tell you that any forecast was accurate. So if your staffing levels go down and the new forecast looks accurate, it looks like you have saved money.
Is a patience factor a wait time input? No. Its just way to convince you you are saving money by scheduling less staff during the periods with the longest wait times.
Question 5 - What about simulations. I’ve hear advanced WFM solutions use simulations. That sounds like they must be pretty in touch with real activity.
So, there are two types of simulations. Most simulations are scheduling simulations. Tens of thousands of schedules are created using random numbers to produce different combinations of agents. Some of those combinations may give better skills coverage and your WFM would choose one of those iterations. However, the root of your wait time problems is your forecast. If the forecast tell you to time your resources poorly, the schedule just operationalizes the bad plan. Hence whether your schedule is built manually, with random number simulations or any other method, you get the same result. If the forecasts are flawed, everything that comes after is destined to fail.
Question 6 - So what about forecast simulations? Some call them detailed event simulations? Don’t they have a more detailed understanding of demand?
Forecast simulations are just another form of loss. Some systems create Loss by having you input a patience factor. Others create loss by studying your historical call abandon rate. Detailed event simulations create loss using random numbers. Keep in mind that any time the word simulation is used it actually means random numbers.
In the case of forecast simulations, the sales pitch is that all calls don’t last exactly the same time. The pseudo-scientific proposal is that this gives you unrealistic wasteful forecast. Supposedly, those forecasts can be made more efficient by simulating call lengths that are randomly different from your average call length. Vendors point to the success of their forecasts by showing comparisons of their forecasts to regular forecasts. Sometimes the simulations are higher. Sometimes they are lower, but the long run average is four to seven percent lower. So they say you save money.
Keep in mind that these are interval forecasts. The only historical inputs are average interval talk times and average interval call counts. The observed forecast differences are a simple bi-product of random numbers and rounding errors. If you produce 100 talk times that are randomly higher and lower than 5 minutes, then the average of those random numbers will vary. Sometimes it will be higher and sometimes it will be lower. Rounding errors ensure that the long run average is between 4% and 7% lower.
As with any loss approach, your staffing levels go down and shifty call counts continue to tell you the forecast accuracy remains high.
Question 7 - But how is it possible to get lower wait times and better service levels without hiring more staff?
If you are relying on interval forecasts it is not possible. The forecast’s perception of demand is frozen to historical call answering rates so the only way to reduce wait times is to hire more staff across the entire day.
SCO forecasts understand wait time problems so intricately that they shift staff across intervals into a better timing.
Let’ take the simplest of any wait time problem. Say just one agent in my call center starts 15 minutes later than he or she should. It makes the call center short staffed for 15 minutes. But once the 15 minutes is over, the shortage does not go away. Its held in queue. So the next interval, every new caller that arrives will have to wait behind the excessive queue that formed in the previous interval. That excessive queue will actually flow across the rest of the day. The timing of agents is so critical that just one fifteen minute shortage can give you long wait times all day long. Move one agent or one break and an day long wait time problem is solved.
The illustration above was the simplest of wait time problems. If you are short by more than one agent or short in more that one interval, you get significant wait time problems cascading across the day.
Interval forecasts don’t understand any of these dynamics. They assume that every interval starts with an empty queue and they consider absolutely no wait time information. Naturally, interval forecasts can’t make the intricate adjustments needed to resolve even the simplest of wait time problems.
The moment SCO removes those limitations, scheduling to that forecast will immediatly shift your staff into an ideal pattern. The SCO forecasts structures your schedule for consistently low wait times and just enough work in queue to keep agents busy. The cost savings, productivity improvements and revenue gains are far beyond any thing you might think is possible. Try it and you’ll see.