CALL TRAFFIC ANALYSIS

Forecasting is as much an art as a science. It involves making projections, not goals and objectives. Forecasts provide an expectation of the future based on facts, historical data, experience, and insight.

Step one is gathering the data., You want stacks of historical data. History is the best indicator of the future. The first place to check is your contact center platform. A robust platform will have historical reports stored for up to a year. If your platform emails or FTP’s reports automatically, many managers save those reports. Seek out those managers. Can you get two years’ worth of data? Much more than two years isn’t relevant with the fast pace of change. Less than 12 months doesn’t represent the complete picture.

Then scour the data for any one-offs., Look for data that looks abnormally low or high, or missing data. Then, we will determine what is triggering this out-of-the-ordinary data. Is that “zero” you see on a holiday and the contact center was closed? Does that holiday change to a different day each year, like July 4th, or is it the same day as Memorial Day, the last Monday in May? For example, if July 4th moves from a Saturday to a Tuesday, how does that impact call volumes on the preceding and the following days? This is the “art” part of forecasting where your experience, intuition and judgment come into play. The data is factual but not consistently representative.

In this next step, you want to “normalize” the data up or down. Let’s say on Thursdays, call volume is easily 5000 calls for the day, but you notice that on the second Thursday it is only 3000 calls. Further research shows that the contact center experienced an outage for four hours that day. Since that’s not a typical Thursday event, you can assume that the contact center would have handled 5000 calls on that day if there hadn't been an outage. The actual number for that day is correct, yet you want to use the typical day of 5000 calls for forecasting purposes.

On the other hand, if the contact center completes mandatory compliance training every third Thursday, then the lower number is accurate and would be the correct number to use for forecasting. If the event is repeatable and predictable, these lower numbers are realistic.
The key is to determine what is causing the data aberration. You must answer that question.

Contact centers MAKE outgoing calls; outbound services make outgoing calls for your organization. 

Examples are:
— Appointment Setting
— Collection Reminders
— Lead Generation
— Insurance Sales
— Market Research
— Mystery Shopping
— Payment Protection
— Event Registration
— Telesales
— Warranty Programs
— Fraud Prevention 

No matter how hard you try to make it an accurate forecast, it won’t be perfect. However, that's a good day if you get 98% accuracy.

It is an estimate, a prediction of the future. Ask yourself how off your forecast is and why. You are likely using a complex method when the simpler ones are more accurate. Complicated methodologies hide or make key assumptions that may not fit your business. Stick to simple techniques.

If you don’t trust the raw data, you can’t trust the forecast. The data quality is proportional to how often you use it, believe it, and correct it. When you use data regularly, you become familiar with its inaccuracies and can get corrections made. When you clean up the data, it becomes a powerful tool for forecasting, and therefore you’re more likely to trust it.

Bias also gets in the way of forecasts., Let’s face it, when you have to make assumptions, it is challenging to eliminate adding some bias. Just be conscious that this happens.
And finally, technology doesn’t make forecasting better. Sound logic in your methodology creates a robust forecast. Technology can make the method more efficient and, thus, more successful.

Contact centers TAKE incoming calls; inbound services take incoming phone calls to your business. This includes:

— Telephone answering service
— Cell phone
— Medical calls

— Overflow/after hours
— Appointment Management
— Building Maintenance
— Customer Service
— Direct Response
— Disaster Response
— Dispatch
— Emergency Call Center Services (911)
— Help Desk
— Hotlines
— Live Chat
— Loyalty Programs
— Order Processing
— Product Recalls

— Virtual Receptionist 

You now have monthly forecasts, but a good contact center manager must also know individual daily predictions. Daily patterns have been the easiest to observe over the past few weeks. You don’t have to go back through the last two years. Select a couple of usual weeks that don’t have any anomalies. Look at the total Monday volume compared to the weekly total. Repeat for Tuesday thru Friday. These percentages reflect your day-of-week patterns.

You can continue this breakdown until you reach half-hour segments. Forecasts must include not only call volume predictions but also handle time predictions. We need the complete picture to calculate workload, predict staffing, and schedule requirements later. Multiply the number of calls by the average handle time. Review handle time predictions to reflect the time of year, day of the week, and time of day since call length can vary for numerous reasons. Understanding and manipulating the raw data into meaningful information can help you make more insightful contact center decisions.

There are three approaches to taking raw data to predict the future:

— Point Estimate is the most straightforward approach. Any point in the future will match the same corresponding point in the past. The first Monday in June last year will be the same as the first Monday in June this year. Of course, anything too simple has its shortcomings. It does not take into account any upward or downward trends.

— Averaging Approaches uses historical data. The most accurate approach is weighted averaging where the most recent events are given more weight than older events. For example, call volumes for the first Monday in July over the past three years are 2400, 2500 and 2600. A simple average would be 2500 calls. A moving average might be 2500, dropping the oldest data, and a weighted average applies 80% to year one, and only 10% on year 3, giving you a prediction of 2570. This approach also misses the upward and downward trends but is most likely closest to the actual forecast.

— Time Series is the recommended approach because it involves time series analysis. This approach considers the effects of trends, season, monthly differences, and historical data. The assumption is that many factors influence the call volume.