~9%
Cost Decrease
$1,683/year in inventory savings.
Forecast-Driven
Inventory Policy
Evaluated five forecasting models on 5 years of daily demand, optimized a continuous-review inventory policy via Monte Carlo simulation, and cut total cost by ~9% while sustaining ≥ 95% service.
≈95%
Service Level
Target met via simulation.
24
Fewer POs/Year
149 → 125 annually.
Simulation Results
Baseline vs. Optimized
10,000 Runs · Grid-Search
| Metric | Baseline R=92 · Q=42 | Optimized R=86 · Q=49 | Delta |
|---|---|---|---|
| Total Cost / Day | $50.94 | $46.33 | -$4.61/day |
| Service Level | 96.8% | 95.0% | -1.8pp (still ≥ 95%) |
| Holding Cost / Day | $30.48 | $29.21 | -$1.27 |
| Ordering Cost / Day | $20.46 | $17.12 | -$3.34 |
| POs / Year | 149 | 125 | -24 |
| Annualized Total | $18,593 / yr | $16,910 / yr | ↓ $1,683 / yr |
Assumptions: Lead time 5 days · Service Z: 1.65 (95%) · Holding: $1/unit/day · Order: $50/PO
Exploration · Tableau
Demand Seasonality & Trend
5 Years · Single SKU
Exploratory analysis across 5 years of daily demand surfaced a clear upward trend, recurring monthly seasonality, and high day-to-day variability.
Methodology
- Five models benchmarked on a train / holdout split (years 1–4 / year 5) using daily demand. HW (multiplicative) was fit on weekly aggregates then disaggregated back to daily via day-of-week factors for a like-for-like comparison — lowest MAE at 3.94 units/day.
- Built continuous-review (R, Q) simulator with proportional Normal noise (σ ≈ 26% of forecast). Baseline policy (R=92, Q=42) derived via EOQ + safety stock formula.
- Grid-searched R ± 20, Q ± 20; selected min-cost policy meeting ≥ 95% service. Validated both policies over 10,000 Monte Carlo runs.