Secrets of Maximizing Forecast Accuracy
The Secrets of Maximizing Forecast Accuracy
This article reveals the secret formula for perfecting the art of high daily, weekly and intraday forecast accuracy. However, you have to read the whole article to understand the secret.
Most who consider themselves to be experts will tell you that high daily and weekly forecast accuracy are important. If you look at the circumstances very rationally, you may recognize that the reverse is true. These accuracy metrics mean almost nothing. It’s not how many calls you answer that matters, it is whether or not you can answer the calls promptly and with consistently low wait times.
High Forecast Accuracy Examples
A simple example will illustrate this. Let’s say you predict 500 calls for a day. 500 calls come in. So you have 100% daily forecast accuracy. What if a large number of callers waited for over 20-minutes? Your forecast accuracy is still 100%. So high daily forecast accuracy says nothing about how well you were prepared to service callers (or how promptly calls were answered). Weekly forecast accuracy is even less meaningful. In the weekly scenario callers could dial, abandon and retry all week long. They may or may not get serviced even if they dial straight through from Monday to Friday. The distress to the callers and the poor planning behind it would be invisible to daily and weekly forecast accuracy.
Now the experts will certainly object to the above criticism of forecast accuracy. They’ll say that if you have high daily forecast accuracy and good average wait times then the two metrics together tell you planning is working out well. However, average wait times are very non-descriptive. A very poorly staffed call center will answer lots of calls instantly (due to overstaffing) then make many callers wait an extraordinary time (due to understaffing). The average wait time may look great even though a large portion of the caller base has experienced conditions that may cause them to rethink their future patronage.
Service levels are even less descriptive. Service levels are unaffected by the longest wait times. If your target is 20% answered in less than n seconds then the 20% of your callers who wait the longest have no influence on the service level. If you are using a commercial WFM solution the software will often over-report service levels by averaging interval service levels to a daily average. For a full explanation of how the over reporting is accomplished, read the following article.
The Importance of Intraday Planning
So if the daily and weekly forecast accuracy mean so little, perhaps you should focus more attention on high intraday forecast accuracy. Unfortunately, doing so will likely result in a great deal of harm to your call center. This is certainly not what the experts will tell you. So how could it be true?
Intraday forecast accuracy is meaningless for two reasons:
1) If you calculate forecasts with offered calls, your forecast will look quite ridiculous. Offered call forecasts are too volatile to use for scheduling purposes. The volatility is the reason that no commercial WFM solution has ever performed offered call forecasting for the past 30 years. The volatility also leaves no hope for intraday forecast accuracy. Before you protest that there are indeed some WFM solutions that do perform offered call forecasting you should know that some ACD’s count calls when calls end. When you count offered calls as they end you are actually counting callers as they leave the system, not as they are offered. Whatever you call it, it’s not an offered call count. It’s actually the count of handled plus abandoned calls.
2) Instead of doing offered call forecasting, the entire industry* has resorted using to answered plus abandoned forecasting. They call this an offered call forecast, but it’s not. The experts will tell you that offered calls is equal to answered and abandoned calls, but it’s not true at all. Answered plus abandoned forecasts provide fantastic intraday forecast accuracy but only because the calculations are entirely fake. The experts never teach this. Based on my experience it’s not because they are trying to mislead. It’s because they don’t actually recognize the fakery is happening. That’s too bad because for the most part they are recommending practices that cripple the ability of a call center to respond to the growing changing needs of callers.
*Note that several WFM vendors perform handled plus abandoned forecasting. They get even higher forecast accuracy and the detrimental effects on the call center are also higher.
Pros and Cons
High intraday forecast accuracy is broadly considered to be a critical key performance management tool. Unfortunately, high intraday forecast accuracy actually prevents call centers from growing and adapting.
The worst thing you can do to a well-functioning call center is to keep calibrating the intraday forecast until you get higher intraday forecast accuracy. The moment you achieve high intraday forecast accuracy you can be certain that you no longer have the ability to answer any more calls than the forecast thinks should be answered.
The Secret Formula
So if your confidence in forecast accuracy remains unshaken and you still want to know how to achieve high forecast accuracy weekly, daily and intraday, here is the secret formula.
Understaff for most of the day, every day of the week. The best way to do this is to pick the minimum answered call count for every interval for the past three weeks and make that your forecast. If the call center manager asks what you are doing, just say it’s a forecast calibration and it’s critical to forecast accuracy – a key performance metric.
Schedule to the forecast and implement rigid adherence monitoring. Make sure agents are heavily penalize for taking calls into their breaks or taking a break early when there are no calls in queue. Insist that supervisors and planners get approval from the planning team prior to any exception. Then, say no to all exceptions. If your call center manager wants to know why, just say high adherence is a key performance metric.
Wait times are going to be really long so it’s best to put a wait time in queue message as a courtesy to callers. When callers hear the message, many will hang up before they are offered to the queue. That’s great because those don’t count as offered calls. Technically, they hung up while listening to the recording. These calls never waited in queue and did not need to be answered. This will greatly enhance your labor productivity, wait times and service levels. Agents will rarely be idle because the call center will be in a constant state of overload. That translates into high labor productivity. Every time the queue starts to accumulate callers, the wait time in queue message will discourage callers from waiting to do business with you. This is a great way to manage wait times and service levels. If your call center manager questions the use of a wait time in queue message, just say it’s an industry best practice.
Prepare each forecast using answered plus abandoned calls. This will ensure that the forecast cannot respond to any increase in demand. This is critical. High forecast accuracy requires understaffing throughout most of the day.
Calculate forecast accuracy using answered plus abandoned call counts. This will ensure that your forecast accuracy is always high. If you forecast 100 calls in an interval, staff only enough agents to answer 100 calls . Make sure they adhere rigidly to their schedules. When you calculate forecast accuracy the offered calls that you projected will be almost equal to the number of calls that you answer. Forecast accuracy will be roughly equal to 100 percent less your abandon rate. Luckily the wait time in queue message will chase away most callers before they abandon.
The last thing you want is too many abandoned calls. This would reduce your forecast accuracy. Callers are stubborn so unfortunately, some will listen to the entire wait time in queue message. Once in the queue they’ll realize that their call really is not going to be answered quickly. That drives higher abandon rates. So the best way to reduce abandons is to establish a healthy short abandon threshold. The longer the better. Some call centers use 10 seconds. But it is better to use 30 seconds. Anyone who hangs up before the threshold does not get counted. This should eliminate the need to count up to 90 percent of callers who abandon. The abandon rate stays low and forecast accuracy stays high so it’s a double win for key performance metrics. If the call center manager asks about the change, just show him the great metrics. The metrics will confirm that you are an expert and that you know what you are doing.
Service Levels might drop a bit when you make all these changes so it’s best to take charge of service level reporting yourself. Produce some really fancy looking reports so that no will have a reason to look at service level reports from the ACD. Now take the interval service levels from the ACD and average them into your own daily service level. These numbers will be much higher than the real daily service levels from the ACD. Now pick the quietest intervals in the day and overstaff them. You want 100% service levels in those intervals to boost the daily average. Overnight periods are perfect for over-staffing. Answer just a few calls instantly and your daily service levels will sky rocket. If someone notices you are not properly weighting service levels, just tell them it’s too hard to change it. Alternatively, tell them you will change it someday. Never change it.
Despite your best efforts to achieve forecast accuracy, there will be days when callers try so persistently to reach an agent that your forecast accuracy will drop. When that happens quickly blame the agents and supervisors for their poor schedule adherence or for having an adherence target that is not high enough.
Always re-forecast after each schedule run. Now on the surface there is no good reason to re-forecast after scheduling. If you have a forecast and you schedule to it–why would the act of scheduling create an immediate need to revise the forecast? Indeed, there is no rational basis to do so.
Luckilly, most call centre managers don’t know that. This is a huge opportunity to perfect your forecast accuracy. Start with a preliminary forecast based on answered and abandoned calls. Now schedule to that forecast. Any imperfections in your schedule will leave gaps between the forecasted number of calls in each inteval and the number of calls that the scheduled quantity of agents should be able to answer.
For example, lets say your 12:00 to 12:15 interval forecast is for 300 calls. Perhaps your schedule is only able to put enough agents in place to answer 225 calls. If you leave things this way, your forecast accuracy will suffer and the credibility of your forecasts might be called into question. Simply change the forecast from the 300 originally forecast to the 225 that you know the scheduled agents will be able to answer.
If you apply this type of change in every interval then you get two amazing benefits.
Firstly, the scheduling algorithm will appear to have been incredibly accurate because it will match the re-forecasted numbers perfectly. People will see the schedule track perfectly to the forecast and say “Wow, that is a an incredibly efficient schedule”.
The second benefit is phenomenal forecast accuracy. That’s because you will have successfully removed all of the influence that demand could potentially have on your forecast. Re-forecasting after scheduling produces a pure capacity based forecast. It is the pinnacle of success for any planner who likes to report high forecast accuracy.
Your forecast accuracy and your schedule accuracy will match that achieved by the Market Leading WFM vendors.
You also want to make sure that no one on the planning team is ever able to schedule to a forecast without immediately re-forecasting. If they were able to skip the re-forecast, someone might catch on that you are taking some liberties with the data.
To avoid this risk, just bundle scheduling and forecasting into a single step. If anyone ever asks why, just tell them that you are mirroring the level of process automation that you have observed in market leading WFM software.
If you perfect these 9 steps you call center will excel at all of its key performance metrics.
Eventually, you might decide to offer these services to more than just your own call center. If so, establish a new WFM company. Attract customers by extolling the virtues of your high forecast accuracy. Explain that the high forecast accuracy is only possible using proprietary advanced forecasting methods. Tell them that your software uses sophisticated simulations to plan for complex multi-skill staffing challenges that are too complex to calculate manually. Encourage each customer to follow your best practices and strive to hit your recommended key performance indicators.
Lost and Found
If this article has indeed shaken your faith in Forecast Accuracy, you may be feeling a sense of lost purpose. Forecast Accuracy was your compass. Does finding out that the compass is broken make you lost? Not at all. Take your eyes off the compass and you’ll see the sun rising in the east.
Good call centers require good strategy. Counter-productive principals like forecast accuracy interfere with proper strategy. The strategic challenges faced by call centers are limitless in variation.
Just a few examples will demonstrate that real call center strategy flourishes when the call center free themselves from strategy killing metrics.
Real Product Support Strategies
Consider a product support call center that is overwhelmed with long wait times, angry callers and low resolution rates. Forecast Accuracy principals would drive that call center towards more punitive adherence and less flexible call answering capacity. Eliminate that misguidance and decision makers start to look at what’s driving the imbalance and how to fix it. Perhaps poor product quality is driving the high need for customer support. Documentation quality might be having the same effect. It may also be that hiring or training have not kept pace with the business growth or the expansion of the product line. Agents and supervisors may actually know the answers to these questions if only the ears of decision makers were open to more than shallow and misleading metrics like adherence and forecast accuracy. Managers only ask for high forecast accuracy because they think it helps. What they really want is for the call centers to support and enable the growth of the business. That’s easier to do when the performance metrics stop strangling the real objectives.
The Real Strategic Analyst
Consider the Analyst who earns a performance bonus for high daily or weekly forecast accuracy. Not only do they have no control over the forecast accuracy, the achievement or non-achievement of a target accuracy rate has no relevance to how timely customers are served. That analyst’s time can be far better spent on matters of real strategic importance. Listen to the longest calls. Listen to the shortest calls. Listen to the calls of the agents with the best and worst first call resolution rates. Analyze those and produce meaningful strategies that enhance the satisfaction of the detailed needs of customers.
Real Blended Call Center Strategies
Consider a call center who’s volume it dependent on product sales. The sales department closes a sale. The call center needs to respond to the follow up requests to dispatch an installer or remotely configure equipment.
The inbound call volume is driven by the quantities of recent orders. Telling the call center to expect 5000 calls this week is of no benefit. They need to know how to time themselves to succeed. Better still, they need a blend of outbound calls and forms processing work to keep them busy while they wait for the unpredictable call spikes to unfold.
If management wants to see high labor productivity, you could achieve that with a bad forecast and high adherence. But that’s counter-productive. Agents will be busy all day long but customer will experience exasperating wait times. Sales will be lost, agents will be demoralized. The metric of high occupancy conceals the large scale loss of real productivity and profitability.
The more intelligent path to productivity requires real strategies to blend inbound work with discretional workloads.
Call Centers will never explore real strategy improvements as long as the managers, supervisors and agents focus on following a broken compass that never moves. Being nimble means having a swift, intelligent and appropriate reaction to changing conditions. Forecast accuracy always works in direct opposition to nimble strategy and responses.
Perhaps you are one of the elite few who are so cognitively nimble that they can change their mind about the status quo for metrics. If so, congratulations on this moment and to the new successes it will soon bring to your business.
Perhaps you are looking for the nimble tools that will guide your call center with truthful metrics that drive real strategies and genuine optimization. If so, please explore the balance of articles on this website.