ops_capacity_planning v1.0.0

name: ops_capacity_planning description: Size an ops team that handles queued work. Computes required staffing with Erlang C, applies occupancy caps and shrinkage, and validates the plan with a queue simulation. Use when deciding how many agents, support staff, or moderators to schedule against forecast volume.

Purpose

Turn a volume forecast into a defensible headcount plan. Demand math first (how much work is arriving), then a staffing model (Erlang C for queue probability), then a simulation check that accounts for caller abandonment, then shrinkage to convert seated agents into scheduled people.

Workflow

1. Demand math. Gather per-interval inputs: forecast contacts, average handle time (talk + hold + after-call work), interval length, service-level goal (e.g. 80/20), occupancy cap, and shrinkage. Use the busiest interval, not the daily average.
2. Staffing model. Run bin/erlang_c.py with those inputs. It returns the minimum agents that hit the service level inside the occupancy cap, plus Erlang traffic, wait probability, average speed of answer, and scheduled headcount.
3. Simulation check. Run bin/queue_sim.py at the recommended staffing with a realistic patience setting. If abandonment is high, the Erlang C plan is understaffed on served demand; add agents until the simulation's service level and abandonment look acceptable.
4. Report the plan. Interval staffing, scheduled headcount after shrinkage, and the assumptions behind both. If the plan doesn't hit the goal, show the trade: agents added, occupancy accepted, or service-level goal relaxed.

Output Contract

• A staffing table: interval, forecast contacts, AHT, agents on queue, scheduled headcount.
• Every number traceable: traffic in Erlangs, service-level math used, occupancy cap applied, shrinkage percentage and source (planned vs measured).
• Caveats stated: what Erlang C assumes (no abandonment, steady arrivals, exponential handle times) and what the simulation covered instead.

Operating Rules

1. Staff to the peak interval. Averages understate the busy hour and the plan fails when it matters.
2. AHT must include after-call work. Trimming AHT to make the math pretty is how understaffing happens.
3. Shrinkage math is required / (1 - shrinkage), never required x (1 + shrinkage). The trap is classic and expensive.
4. Cap occupancy around 80-85%. Sustained higher burns out agents and quietly degrades handle time, which feeds back into more load.
5. When Erlang C and the simulation disagree, trust the simulation for what callers actually experience and trust Erlang C as the defensible baseline for the budget conversation.
6. Every factual claim in this skill's outputs (benchmarks, typical percentages) must be verified against a source. Do not invent shrinkage or occupancy norms; measure them or cite them.