Forecasting trends, advanced technology, precise returns.
Call forecasting is what helps us operate call centers without any issues. It provides us with a clear view of call traffic changes related to peak hours and off-peak timings, assisting us in making informed scheduling and staffing decisions. Although call forecasting might be challenging, it allows our planners to determine the ideal accuracy rate for future projections. A poor projection results in a very appalling management of call volume and accurate forecasting is the foundation of the WFM team’s quality performance. At ChatPandas, we never fail to account for the amount of calls, emails, contacts overall, workload, (Full-time equivalent) FTE, or any other factor impacting the efficiency of our traffic flow.
How do we calculate Forecast Accuracy?
We continuously measure all types of traffic in comparison with the Average Handling Time for an always-on-the-go call answering system. If we only estimate the volume of traffic without doing any study of how much time each agent takes to handle a single call, the prediction will be completely inaccurate. The following KPIs are important for keeping a call center dashboard organized:
Standardized Interval
We relatively have some freedom to modify this metric in accordance with your company’s needs but still we keep 30 minutes as our standard unit of measurement. This parameter can be measured on a weekly or monthly basis, but as we all know things can happen in a call center in a few seconds, so we don’t wait for weeks and months to measure results.
Ground Volume
It is difficult to measure an actual volume perfectly since all incoming call data, including repeat calls, abandoned calls, and answered or resolved calls are all included in the list of incoming calls. It’difficult to program the software in such a way to separate resolved cases from the unresolved. Thus instead of monitoring calls that are answered, we conduct a careful examination of all incoming call. This strategy may seem far from perfect and a lot of work, but it’s still better.
Overhung Call Volume
Calls that connect in one period of time but linger until the next are referred to as overhung calls. The occupancy rate rises with an increase in call volume and decreases with a reduction in the call volume. Thus, we do not compute the intervals that are smaller than twice our Average Handling Time. In this manner, measurement is based on an excessive number of agents taking calls from the prior interval. Precise accuracy helps us to train the employees in accordance with the expected flood of incoming calls.
Defining Percentage Error
It is a relatively straightforward metric also known as percent difference. The gap between the actual volume and the anticipated volume is how our analytics calculate Defining Percentage Error. Its formula is as follows:
Percentage Error = Actual Volume – Forecast Volume/Actual Volume x 100
Our forecasters employ a reliable formula by determining the average volume of various 30-minute intervals at particular times of the day.
Mean Absolute Percent Error (MAPE)
Just monitoring the percentage of average volume of calls, or calls offered, is how we calculate Mean Absolute Percent Error. MAPE is a helpful technique for percentage representations of large numbers since it makes their understanding very simple. Using this formula, we can simply determine the percentage inaccuracy of each period:
Mean Absolute Deviation (MAD)
We calculate the mean absolute deviation using error percentage, which is a very straightforward metric. MAD measurement is more appropriate for small firms. It goes along with figuring out the mean average of the errors (or deviations) for the given data.
Standard Deviation Rate
Analyzing a multitude of data while measuring the accuracy rate over an extended period is beneficial. To do this, we compute the standard deviation of the variation percentages. It is one of the most helpful tools used by forecasters since it displays even mean variations between the rise and fall of longer, average variations. Its formula is as follows:
Standard Deviation (s) = Standard Error * √n
Variance = s raise to power 2
Correlation Coefficient Rate
Correlation coefficient rate helps when we compare the deviations from one interval to another. It allows us to compare the call frequency of two different times in order to determine whether or not their patterns are similar. For instance, comparing the call volume on one Tuesday to another Tuesday can points us in the direction of effective workload arrival numbers and AHT variation measurements. This analysis enables our forecasters to determine whether or not what occurred on that Tuesday was beneficial. The formula for Correlation Coefficient Rate is as follows:
How do we Optimize Forecast Accuracy?
Several methods can be used in order to measure forecast accuracy and implement its results. We implement the following strategies for increased accuracy:
Stay Focused
Our analytics take into account the workforce management because they have an impact on prediction accuracy. By finding out the causes, they draw attention to irregularities in the workload (i.e. marketing campaigns, mailings, billing cycles, ticket management). The coordination of all departments contributes to a more uniform call frequency rate.
Accuracy Rate Analysis
Instead of just focusing on an increase in volume, our goal is to ensure accuracy for our valuable clients. We choose an accuracy rate analysis that is appropriate for the industry or customer interaction. By making daily analysis as our standard, we keep the prediction accuracy as straightforward as we can.
Focus on Time
We create a case for greater or lower demand once we are aware of the number of agents required at a given moment. We use the +/- 10% FTE general rule of thumb for it. It enables us to quickly add or subtract people to account for discrepancies between our prediction and what actually occurs.
Demand Forecasting
We never disregard the significance of prediction accuracy based on items with a longer shelf life (such fresh food and medications) in comparison to those that are slower moving (the merchandise with less customer demand within a specific time interval). We deploy the agents in line with the demand urgency criteria on a significant demand base forecast. Depending on the chosen metric and how calculations are made, the collection of data might produce great or awful for forecast accuracy measurement. Our analytics keep doing modifications in it to better fit our call traffic.