Secrets to High Scheduling Accuracy
The Secrets to High Scheduling Accuracy
Very experienced WFM consultants will typically tell you that there is some “secret sauce” of formulas that allows more advanced WFM solutions to achieve higher scheduling accuracy (compared to “less advanced solutions”). The consultants don’t know what the formulas are. They have only witnessed the higher scheduling accuracy. So by experience, they can point you to the solutions that must be more advanced.
Today you will trump the knowledge of the experts by learning the exact list of ingredients and the the secrets of how they work together.
At the end of this article, we tell you the six ingredients that making any scheduling algorithm produce phenomenally high scheduling accuracy.
But don’t skip to the end. You need to learn the secrets of scheduling accuracy first. Otherwise you won’t know what to do with the ingredients.
These are the topics that you are about to master beyond the comprehension of the most experienced WFM consultants:
- What is scheduling accuracy?
- When is it important?
- Is it hard to do?
- What is the best way to maximize it?
What is Scheduling Accuracy?
Scheduling accuracy is a comparison between the forecast that you are trying to satisfy in each interval and the number of staff that end up being planned for each interval. In theory, a good fit between schedule and forecast makes your call center more prepared to service customers.
When is Scheduling Accuracy Important to Your Business?
Scheduling accuracy is only important if the forecast is sound. Interval based forecasts are rarely sound for two reasons:
- They have no concept of historical wait times because the only two historical inputs are talk times and call counts. You get the same forecast no matter how long callers have been waiting.
- Interval based forecasts are derived from capacity based measurements that ignore the vast majority of demand indicators. Interval based WFM solutions rely on capacity based measurements to produce an illusion of forecast accuracy and to stabilize the forecast around established staffing norms.
There are roughly 10 tricks that many WFM solutions use to exaggerate forecast accuracy and prevent schedule movement. These 10 tricks are explained fully in the following article:
If you understand the art of manipulating data to produce high forecast accuracy, you are in a good position to understand that scheduling to a bad forecast is not a good thing for your business. Scheduling accurately to a bad forecast is like driving accurately through a mine field – with the objective of hitting each mine.
Scheduling accuracy is only important to the success of your business if you are starting with a forecast that takes you out of the mine field.
When is Scheduling Accuracy Important to Your WFM Vendor?
For most WFM vendors, scheduling accuracy is always very important. There are two benefits for the vendor:
- Accurate schedules are part of the stabilization cycle. The stability of the forecast erodes if the schedule can’t complete the cycle with another round of staffing levels that make the next forecast come true. Any drift in the schedule causes drift in the next forecast and so forth.
- High schedule accuracy is something that WFM vendors like to have bragging rights to. Certain vendors claim to have higher accuracy than their counterparts. In support of these claims, you will see a lot of very experienced WFM consultants observing that their favorite WFM solutions do consistently produce higher scheduling accuracy. What the experts don’t understand is that in the majority of cases, scheduling accuracy is just a clever illusion.
To understand when it is an illusion, you need to first understand what can make scheduling accuracy hard, or easy.
Is it Easy or Hard to Achieve High Scheduling Accuracy?
Scheduling accuracy is the very simplistic measurement of the overage/underage in each interval. Scheduling accuracy gives no consideration to qualitative outcomes including the degree of inconsistency of the starts, ends and breaks for each agent from day to day and week to week. Poor quality schedules are hard on agents but may have a numerically high scheduling efficiency.
If you ignore the quality of the schedule, creating a high efficiency schedule is very easy. All you need to do is work from left to right, handing out enough start times to keep pace with the forecast. Then you shuffle the breaks around so as to minimize the gap between the forecast and schedule. It’s a little bit more complicated than that — but not much.
The majority of WFM solutions including all three of the market leaders use random numbers to produce lots of random schedules (10,000 to 100,000). They then evaluate those schedules to identify the one that offers the closest fit to the forecast. A smaller number of vendors use logic based methods and have no reliance on random numbers.
While the two different approaches are likely to produce different outcomes for quality, finding a solution of best numerical fit is extraordinarily easy. So much so that given an identical forecast everyone’s scheduling accuracy should be pretty much the same.
If the forecast is very jagged with many steep spikes and sudden valleys, then everyone’s scheduling accuracy will be relatively low. That’s because moving breaks from spikes to valleys has limits. Once all the breaks are pushed out of a spike, the only way close a gap is to add a shift. When you chase a spike with more staff, you create overages on either side of the spike. Overages and underage’s become unavoidable.
If the forecast is very smooth then scheduling accuracy becomes very easy for everyone’s scheduling algorithm.
What is the best way to Maximize Scheduling Accuracy?
Winning at scheduling accuracy is a lot like a card game. The best way to guarantee a win is to switch the deck when no one is looking. And that’s exactly what happens with the solutions that claim to have the highest scheduling accuracy. They “mush” jagged forecasts into smooth ones. Then, they attribute the higher scheduling accuracy to their secret sauce of more advanced algorithms and simulations.
Unless you prepare your own forecast and compare its eccentricities to the relative smoothness of a secret sauce forecast — you are unlikely to notice that the deck has been switched.
The same is true if you compare the scheduling accuracy of any two WFM solution. Of the available tricks to smooth a forecast, some vendor use most or even all of the tricks. The vendor that uses more forecast smoothing tricks will always score highest on scheduling accuracy.
The experts that have aligned themselves with certain WFM solutions will disagree. They will insist they have done comparisons. But for any comparison to be valid you would have to force two different WFM solutions to arrive at exactly the same forecast before they schedule. That just does not happen — thanks to the secret sauce.
Where is the Secret Sauce of Maximizing Scheduling Accuracy?
It turns out there is a 50% overlap between the tricks that WFM vendors can use to maximize forecasting accuracy and the tricks they can use to maximize Scheduling Accuracy.
Of the ten or so techniques that create an illusion of high forecast accuracy, six of them are very effective at smoothing the forecast. The byproduct of smoother forecasts is high scheduling accuracy. Hence the WFM vendors that use most of the forecast accuracy tricks also have the smoothest forecasts and the highest scheduling accuracy.
Do the Tricks Have Consequences?
The tricks are not harmless. Forecasts were never meant to be smooth. They should reflect real fluctuations in historical demand so that staff can be times in ways that provide consistently low wait times.
The moment you smooth a forecast the effectiveness of the plan falls apart. As the day progresses, the first spike in demand is under represented by a smoothed forecast. That produces a swelling in the queue that will rip across the rest of the day. Many more real spikes and valleys will unfold. The smoothed forecast will never prepare the call center properly for any of the real demand activity.
For evidence of this look to the call centers of the airlines, cellular providers and ISP’s that rely on leading WFM solutions. Have you experienced what you would call consistently low wait times?
The impact of ineffective schedules is typically lost on the most experienced call center planners, managers and consultants. It all gets overlooked for the same reasons that a card maverick is able to switch the deck. The maverick creates a distraction before the switch. In the WFM industry the distraction is the metrics:
- High forecast accuracy
- High scheduling accuracy
- Over reported daily service levels and daily wait times (using averages of averages)
- Under reported abandons (using short abandon thresholds)
Illusions are easy when everyone is looking in the other direction. Call center managers, planners and consultants need to be looking in the right direction. The revenue, profitability and growth of every business depends on it.
How Not To Be Fooled
This article has an agenda. However, there are no tricks. You are not being hypnotized into buying one solution over another. The only agenda is to expose the truth. How you use that truth is at your discretion.
You can use it to see through illusions and purchase based on price and features without fear of missing out on some secret sauce. You can elect to continue to operate in the realm of interval based planning, but with a higher awareness of which metrics have been tampered with.
Another option is to walk away from the 100 year old game of capacity based interval planning and step bravely into the modern world of truthful forecasting and scheduling. If that’s your choice, this web site tells you how to get there with great ease and amazing results.
If you work with one of the many market entrant interval based WFM solutions and you want to learn the secret sauce of phenomenal scheduling accuracy, here it is…
Secret Sauce Ingredients:
Here are the six most effective tricks used to enhance scheduling accuracy.
1) Smooth the forecast.
Literally, it’s that simple. Just use arithmetic smoothing to even out the spikes and valleys. The scheduling efficiency will automatically appear higher because the forecast’s propensity to jump above and below the staffing curve will have been purged. It’s a really easy forecast to schedule to because you will have removed the inconvenient fluctuations. When you schedule to a smooth forecast, the resulting smooth schedule provides a muted response to demand. The sluggish staffing curve can drive long queues that may ripple across the day. To mitigate these risks, make sure that you use all the available tricks to over report service levels, under report wait times and under report abandons.
2) Smooth the Call Counts Even More.
Instead of count calls as they are leaving the queue (answered plus abandoned), count them as they leave the system (Handled plus abandoned). Either method will smooth the forecast because the arrival rate of calls is discarded in favour of the much smoother rate at which calls exit. Calls exit the queue more smoothly than they arrive in the queue. Calls exit the system more smoothly than they exit the queue. Thus you get a smoother forecast from counting calls as they leave the system. You also get longer wait times. The longer queues may be uncomfortable for callers but it ensures that the planned answering rate is never interrupted by a lull in demand. Any time you ignore demand in favor of capacity the outcomes are smoother, and easier to schedule to.
3) Smooth the forecast to the schedule.
Do this by regulating call counting to the capacity that was scheduled. If 100 calls suddenly jump into queue, the demand spike won’t create a spike in the forecast because you will only count the calls at the smooth rate that they leave the queue (or system). This enhances forecast accuracy because it makes it very difficult to count more calls than the call center was scheduled to answer (or finish). Schedule accuracy is higher not only because the forecast is smoother but because the forecast is smoothed in a very particular way that tracks to the day’s staffing levels. If next week’s forecast asks for a similar number of staff in each interval as the scheduling engine was able to schedule this week, high schedule accuracy for the forthcoming week is inevitable. Naturally there is a three week average that goes into the forecast. That averaging introduces another layer of forecast smoothing.
4) Ignore short abandons and use a long “short abandon threshold”. The more abandoned calls that can be swept under the carpet, the closer the call counting will track to the call answering rate. The abandon rate has more potential volatility than the answering rate so squeezing out the abandons produced a smoother forecast. You get higher forecast accuracy and higher scheduling accuracy.
5) Rigidly enforce agent adherence.
This will ensure that agents don’t answer more calls than the forecast calls for. The stronger the adherence, the smoother the forecast. A highly regulated call answering rate can’t respond to demand fluctuations. When you hold the call answering rate to the planned levels it ensures that the call counts that go into the next forecast are very smooth and very consistent with what the scheduling engine has historically staffed. Naturally forecast accuracy is also high because the call counting rate is held to the level that was forecasted.
6) Secretly replace the forecast after each schedule is created.
Start with a preliminary smoothed forecast that has relatively little influence from call arrival rates. Schedule to the smoothed forecast. Now delete the preliminary forecast and replace it with one that is reverse engineered from the staffing levels. If 2:15 – 2:30 gets 100 agents on the phone and calls last 5 minutes then forecast 300 calls.
Some WFM solutions actually do this. By default, forecasting and scheduling is integrated the into a single step. Some have done the same for schedule re-optimization.
You can test this by creating two schedules that differ only in the number of agents available. Start with ample agents and create a “Forecast & Schedule”. Take a copy of the forecast. Now create a new schedule with not nearly enough agents. Put half of them on stress leave. Then create a new “Forecast & Schedule”. Finally, compare forecasts.
Are the forecasts are different? If they are, the reason is that forecasts are being reverse engineered from how many agents could be scheduled. That’s a pure capacity based forecast and it’s the most toxic form of planning. On the upside, the scheduling accuracy appears perfect because the forecast is always replaced with one that matches the schedule.
That’s it. Those are the ingredients to the secret sauce of claiming that your WFM solution schedules more accurately than the competition.