Forecasting Workload

Forecasting Workload

Forecasting workload is a form of predictable analytics, we are using historical data to anticipate the future.

Forecasts may vary in frequency based on operational requirements, but as a rule of thumb, the following schedule should enable effective planning and support adjustments close to the contact arrival date. Forecasts are provided for all live agent channels: voice, chat, email/mail, social media intervention or listening posts. It is also important to note, that there are two best practices for forecast accuracy dependent upon the client’s operating model:+/- 3% margin of error required by interval (exception during midnight hours) if there is little elasticity in the workforce. With little access to increasing or decreasing agent levels intra-day, it is important that the forecast be highly accurate. Generally it will be constructed using multiple econometric models external to the WFM tool.+/- 5% margin of error average for the day (exception midnight hours) if the operation generally staffs at 10% or higher above required levels. This use of an agent buffer provides enormous flexibility allowing the operations to both scale down (voluntary time off offered intra-day) and scale up (retain the buffer rather than offer voluntary time off for 10% additional coverage). A client’s staffing plan, the ability to access resources during peak periods that provide flexibility in staffing up for peak loads makes forecast accuracy less critical and places a greater burden on the intra-day management team. Forecasting approach normally recommended as a best practice by McIntosh but obviously available to be customized to a client’s specific environment:

  • Rolling twelve month workload forecast (contact volumes and transaction times) updated monthly with volumes forecasted by month and by week.The rolling twelve month forecast is used for annual budget planning and building the annual capacity plan by month for workstation and technology requirements and agent level staffing. Workload is forecasted at the business unit or functional level and then further refined for a skill-set (alignment of call segments with agent skills) forecast that drives the staffing plan.
  • Rolling twelve week workload forecast updated weekly with workload volumes and transaction times forecasted by week, by day, and by half hour interval. The twelve week forecast is used for resource planning and identifies near-term requirements for increasing/decreasing staffing levels and is leveraged by HR to update the hiring/training plan. Scheduling should act upon this plan to modify schedules in advance of overstaffed/understaffed dates. Depending upon new hire training duration, some organizations may benefit from using a sixteen week or longer rolling forecast.
  • Rolling three week forecast updated daily by day and by half hour interval. This forecast is released to the Intra-Day Management team for final adjustments to the forecast and to staffing levels based on near-term trends or new information that becomes available. Decisions for advance leave or solicitation of overtime or extra hours is based on this forecast. Intra-Day may assume control of the forecast anywhere from three to seven days prior to volume arrival.
  • Special forecasts for holiday workload (day before, day of, day after) which drives holiday shift bids that are released four weeks before the holiday.

After forecasting the planned workload (contact volumes and transaction times) by channel and by product or business unit and skill set, staffing plans are developed using the interval level call forecasts. Generally applications such as NICE’s IEX or Aspect’s eWFM leverage historical arrival patterns by day of week, week of month and apply statistical methods such Erlang theory to identify interval staffing requirements from which schedules can then be developed. It is important to note, that McIntosh advocates using an econometric model for forecasting annual, monthly, and daily volumes with interval level allocation of the forecasted volume accomplished through the WFM tool.

The contact volume forecast also is used to plan facility and technology requirements. The planning exercise should be reviewed on a quarterly basis for physical assets (site, workstation, training rooms, voice and data capacity) with staffing plans reviewed weekly. Staffing plans are reviewed more often due to the operations ability to modify schedules and adjust staffing levels by day or within days. As a best practice, McIntosh recommends a daily debriefing of the prior day’s performance relative to service level attainment and agent productivity with all intervals outside the target KPI to be assessed and the root cause of the gap identified. Based upon this analysis, forecasting might adjust the model, scheduling might adjust assumptions they provide to forecasting, and intra-day may recognize an urgent need for vigilance in the near-term.