How to forecast staffing needs from demand to schedule

For task-based work, use three connected calculations. Workload hours equal forecast demand multiplied by average task time. Productive staff equal workload minutes divided by interval minutes and target occupancy. Scheduled staff equal productive staff divided by one minus shrinkage.

For queued work with a response-time target, such as calls or live chat, simple division is not enough. Use Erlang C or Erlang A after the demand forecast because arrival patterns, waiting time, and abandonment change the number of people required.

Start with historical demand at the level where staffing decisions happen. A service desk may use tickets by hour. A contact center may use calls or chats every 15 or 30 minutes. A field or project team may use jobs or work hours by day or week.

Clean obvious data errors, but do not remove real peaks only because they are inconvenient. Add known information such as holidays, campaigns, product launches, opening hours, and seasonal events. Keep a baseline forecast and record each manual adjustment so you can later test whether it helped.

Demand counts items. Staffing covers time. Convert calls, tickets, orders, visits, or tasks into workload with a recent and representative average handling or task time. Use separate times for work types that need different skills or take materially different amounts of time.

Occupancy protects people from a plan that assumes every paid minute can handle demand. Shrinkage covers paid time that is unavailable for the planned workload, such as leave, sickness, meetings, training, coaching, and breaks. Service-level rules then account for how quickly queued work must be answered.

Keep each assumption visible. A staffing requirement is much easier to review when a manager can see whether a change came from demand, handling time, occupancy, shrinkage, or the service target.

A requirement says how many people and skills are needed by interval. The schedule must place real employees against that requirement while respecting availability, contracts, breaks, leave, and shift rules. Compare required coverage with scheduled coverage before publishing the week.

Do not hide a shortage by averaging the whole day. Four extra people at 15:00 do not solve a four-person gap at 10:00. Keep the forecast, requirement, and scheduled coverage aligned at the same interval.

After each schedule cycle, compare forecast demand with actual demand by interval. WAPE gives a volume-weighted error measure, while bias shows whether the forecast tends to run high or low. Review both because a forecast can have an acceptable average error while repeatedly understating peaks.