Act one
Run the same day twice.
This day is planned correctly on average. The forecast is right, the schedule funds it, and people show up as planned. The remaining uncertainty is arrival and handle-time variation inside the day. Run it more than once before judging the plan by one outcome.
View run results as a table
Act two
When the day misses, name which gap it was.
A miss is not one thing. It can begin in the forecast, the plan, absence beyond allowance, or adherence. Attribution does not prove a cause. It identifies which layer warrants the next question. Change one assumption at a time and watch the accounting move.
The model
How much of the forecast requirement the staffing plan scheduled.
The floor
The plan assumes 90%. A lower value means scheduled coverage that existed on paper and not on contacts. Above 90%, the day holds the plan. Extra adherence cannot add capacity the schedule does not contain.
Sick calls, no-shows, and day-of pulls beyond what planned shrinkage already budgeted.
Share of off-plan minutes papered over by editing the schedule after the day started.
The chart shows agent-hours. For service level only, fractional effective staffing is treated as a time-weighted mix of the two adjacent whole-agent Erlang-C states. It is not rounded down.
View chart data as tables
Act three
Metrics are proxies. Proxies need a paired truth.
A measure can improve while a consequence it misses worsens. The three trajectories below are illustrative stress stories, not data or causal estimates. Their job is to make the pairing explicit before someone promotes a proxy alone.
View scenario data as a table
The discipline is metric pairings. Do not promote a proxy without naming the evidence that could show what it hides.
- Average handle time→repeat contacts and callbacks: speed may create workload elsewhere
- Occupancy→unplanned shrinkage and attrition: compare the pattern. Do not assume the mechanism
- Adherence→schedule settling and coaching delivered: a high score can coexist with a plan that protects neither
- Service level→which gap caused the miss: the average does not supply attribution
Act four
The cost of a headcount no starts with capacity not granted.
Staffing meetings usually end with an FTE request and an FTE grant, not a declared occupancy target. This 16-week stress test holds the model’s workload at the amount the request would carry at a 91% recovery point, then shows what happens under explicitly illustrative feedback assumptions when less capacity is funded.
It is not a forecast that a headcount decision causes absence or attrition. It makes the assumption visible so it can be challenged, measured, or rejected.
View scenario data as a table
The point
Reality is what happens on the floor.
A forecast is a model of the floor. A good one states its assumptions, makes its limits visible, and expects to be wrong in knowable ways. The work is not defending the model. It is naming the gap, choosing the next test, and pairing every promoted metric with the evidence that keeps it honest.