Wil:
I understand your frustration. WFM solutions can in many cases set up a fully automated process that iteratively pulls your forecast into patterns that your own insights tell you are off the mark. For this very reason, we stayed with Excel based forecasting and scheduling for years.
The problem is that most WFM solutions take a pretty superficial and unrealistic view of historical demand. For example, I know that when we are building towards our peak period there is just no-way we should be staffing as if the call demand is random. It is not random, its is accelerating upwards. If we use an Erlang formula then it tells us to staff for relatively flat call volume. If I take the advice, my service levels tank, abandons take off and complaints start pouring in.
Also, if you look at the majority of WFM solutions out there, they will collect past service levels but make no use of them for forecasting purposes. So 300 calls with 5-minute AHT give you the exact same demand for agents whether the historical service levels were fantastic or terrible. Common sense tells us we need to staff-up or staff earlier to improve on poor service levels the next time around. So why invest the time expense and energy in a system that automates really bad decisions that a more thoughtful process would avoid.
I think the senior managers who bought your WFM solution are going to give you a really hard time about not using it for forecasting so you may want to educate yourself on how to defend your position.
What we ultimately did was to wait for a WFM solution that created more thoughtful forecasts. Eventually we found it and its been just fantastic for the past three years. This system uses a more modern approach to forecasting that intricately interprets the real distribution of past events.
Unlike most WFM solutions which only collect call counts and average durations, this one pull in the complete details of each call and evaluates historical demand with precision. This system gives an appropriate staffing level during periods that build towards peak demand. During other periods when calls really are arriving relatively flat the system staff much lower than an Erlang calculation. This works out well because flat call volume is easy to absorb into a service level without the overstaffing buffer introduced by an Erlang calculation. If we had poor service levels, high abandons or long wait times then the system automatically determines how to shift future forecasts and schedules to improve performance. The transformation of our forecast curve was very noticeable when we first implemented the system. Today the adjustments are more subtle. The company has a few different names for their technology but essentially this is “transaction based forecasting” not “interval based forecasting”.
Better forecasts gave us the confidence to fully automate our forecasting and scheduling process. Naturally that has reduced administrative effort but this is just one of many improvements that we have recognized. Agent have benefitted from a broad cross-section of web based tools for shift trading, vacation bidding, time-off requests and of course schedules on-line. As a result, our absenteeism has dropped significantly, especially during the unpopular weekend and evening shifts. The agents also see their own adherence in real time. Equipping the agents to manage their own compliance has made it easy to achieve great service levels consistently.
The entire system is fully web based so it is accessible to agents and supervisors from both home and office.
According to our experience your reluctance to use WFM for forecasting is warranted. As long as you don’t fall victim to the forecasting caveats, the process automation benefits of WFM are definitely worth exploring. However, don’t underestimate the damage that fundamentally flawed forecasts can do. If your staffing patterns get corrupted then the potential benefits of WFM/WFO will most likely slip through your fingers.
Cheers,
Kevin
|