Optimal Headcount Model
A three-stage staffing optimization model that determined the optimal sales-floor configuration across multiple departments — anchored to a profit-based ROI cutoff, refined with inbound-revenue and attrition adjustments, and stress-tested with Monte-Carlo simulation to produce P10/P50/P90 revenue bands for risk-aware planning.
Why it mattered
The sales floor's departments needed an evidence-based way to determine the right staffing configuration — one that captured operational efficiency, productivity, and revenue performance, while eliminating both capacity shortfalls and the cost drag of overstaffing. Headcount decisions had previously rested on a mix of historical precedent and intuition, with limited visibility into which departments were genuinely productive at the margin and which were carrying cost the business couldn't justify.
Built across three model versions
A three-stage staffing optimization model — base optimizer, inbound + attrition-adjusted, and Monte-Carlo simulated — anchored to a profit-based ROI cutoff and refined with leader input across versions.
Base optimizer
A clean staffing delta for each department under steady-state conditions. Defined "productive" using a profit-based ROI cutoff (revenue covering total compensation by a defensible multiple). Built ramp logic to credit new hires fairly during tenure. Produced two diagnostic lenses — a historical benchmark view (average revenue per productive rep) and a current-roster view (productive headcount today). The delta between needed and current headcount surfaced hires versus surplus, department by department.
Inbound + attrition adjustments
Stripped inbound-generated revenue out of the ROI calculation to credit only outbound selling effort, then padded needed headcount to cover attrition — reflecting the real cost of churn rather than steady-state assumptions. Surfaced departments where inbound was masking outbound underperformance.
Monte-Carlo simulation
Ten thousand simulated seasons modeling random attrition and mixed-hire quality, producing P10/P50/P90 revenue bands per department and at the company level. Flagged the statistical nuance that summing department-level percentiles overstates downside risk — percentile-of-a-sum is not the same as sum-of-percentiles unless every department moves in perfect correlation. The company-wide band, derived from the full correlation structure of the simulation, is the trustworthy reading for consolidated planning.
The shape of the work
- 3 model versions — base optimizer, inbound + attrition adjustments, Monte-Carlo simulation
- 10,000 simulated seasons in the Monte-Carlo layer
- 12-month forward planning horizon
- P10 / P50 / P90 — revenue bands for risk-aware headcount decisions
How leaders use it
Supports staffing decisions with consistent, comparable evidence; helps leaders right-size teams and make resource allocation choices with greater confidence.
- Staffing configurationIdentifies which departments are over- or understaffed today, with a "needed headcount" anchored to productivity rather than precedent.
- Risk-aware planningThe Monte-Carlo P10/P50/P90 bands let leaders see the range of plausible revenue outcomes under attrition and hire-quality uncertainty, not just a point estimate.
- Compensation framingBy separating inbound vs. outbound revenue and crediting only productive selling, the model surfaces where comp may be paying for revenue the team didn't generate.
- Goal calibration upstreamThe gap between current and needed headcount feeds into goal-setting conversations, not just hiring — connecting staffing logic to target-setting logic.
What this unlocks
A reusable framework for ongoing optimization — department-level attrition rates can be tuned individually rather than treated as a single global number, inbound revenue logic can be refined as channel attribution improves, and selective ROI thresholds can be set per department to reflect different role economics (inside sales versus premium teams behave differently). The same backbone supports successive planning cycles without rebuilding the simulation layer.