2026 FAU / IMI
SCM Case Study Competition
Enter the password to continue.
Time-Series · Causal Modeling · Scenario Analysis
This study analyzes crude oil pricing dynamics using both time-series forecasting and causal modeling to understand key drivers of price movements and generate reliable forecasts for Brent crude and VLSFO under different market conditions.
Key findings across time-series, causal, and scenario analyses.
Both Brent and VLSFO prices are primarily driven by trend dynamics rather than seasonal patterns. Holt's Trend model performs best for Brent, while Moving Average performs best for VLSFO by error metrics.
Industrial production and geopolitical risk are the most significant drivers across both markets. Inventory changes consistently reduce prices. Brent is more sensitive to macro-financial variables (USD Index), while VLSFO is more influenced by demand and freight factors.
Demand shocks create the largest price movements. Combined shocks confirm that demand dominates even in crisis conditions. VLSFO exhibits significantly larger responses than Brent across all scenarios.
Both Brent and VLSFO forecasts indicate a gradual downward trend through mid-2026. Brent is forecast to decline from ~$64 to ~$61, while VLSFO is forecast to decline from ~$390 to ~$382 (Holt Trend model).
Train/test split: July 2018–December 2024 (training) · January 2025–January 2026 (testing). Models evaluated on RMSE and MAE.
Brent exhibits strong trend behavior with significant volatility during 2020 (COVID collapse) and 2022 (Ukraine war spike to ~$120). No consistent seasonal pattern is present. The series is non-stationary in levels but becomes stationary after first differencing.
Price movements are primarily driven by evolving market conditions rather than stable long-term averages, indicating underlying trend persistence.
| Model | RMSE | MAE |
|---|---|---|
| Holt Trend ⭐ Best | 2.9897 | 2.4189 |
| Moving Average | 4.0832 | 3.4611 |
| SES | 6.9886 | 6.1682 |
| SARIMA | 7.1992 | 6.3623 |
| ARIMA | 7.1992 | 6.3623 |
| HW Seasonal | 10.9149 | 10.2594 |
| HW Trend+Seasonal | 11.4528 | 10.7576 |
Gradual downward trend reflecting recent price softening. Both commodities show similar directional trends.
| Date | Forecast (USD/bbl) |
|---|---|
| 2026-02-01 | $64.23 |
| 2026-03-01 | $63.49 |
| 2026-04-01 | $62.76 |
| 2026-05-01 | $62.03 |
| 2026-06-01 | $61.30 |
| 2026-07-01 | $60.57 |
VLSFO shows larger fluctuations in magnitude compared to Brent, indicating higher sensitivity to market shocks. The 2022 spike reached ~$830/MT. No consistent seasonal pattern is present. The series is non-stationary in levels but becomes stationary after first differencing.
| Model | RMSE | MAE |
|---|---|---|
| Moving Average ⭐ Best (accuracy) | 32.5401 | 28.2309 |
| Holt Trend (Final Forecast Used) | 58.6174 | 49.9511 |
| SES | 72.0187 | 61.9386 |
| SARIMA | 78.8284 | 68.8409 |
| ARIMA | 78.8284 | 68.8409 |
| HW Seasonal | 95.1496 | 88.6001 |
| HW Trend+Seasonal | 96.2378 | 89.5946 |
VLSFO exhibits larger absolute price changes than Brent, reinforcing its higher sensitivity to market conditions.
| Date | Forecast (USD/MT) |
|---|---|
| 2026-02-01 | $390.42 |
| 2026-03-01 | $388.64 |
| 2026-04-01 | $386.87 |
| 2026-05-01 | $385.09 |
| 2026-06-01 | $383.31 |
| 2026-07-01 | $381.54 |
Crude prices are primarily trend-driven, not seasonal. Seasonal models performed poorly for both commodities.
ARIMA and SARIMA did not outperform simpler models, suggesting complex autoregressive structures provide little additional value here.
VLSFO demonstrates more short-term variability, making it more responsive to recent price movements than Brent's smoother trend dynamics.
Multiple linear regression with lagged independent variables to quantify the impact of macroeconomic, supply, and freight/risk factors on Brent and VLSFO prices.
| Model | R² | Adj. R² |
|---|---|---|
| Full Model ⭐ | 0.602 | 0.568 |
| Freight/Risk Model | 0.431 | 0.412 |
| Macro Model | 0.231 | 0.214 |
| Supply Model | 0.177 | 0.159 |
| Model | R² | Adj. R² |
|---|---|---|
| Full Model ⭐ | 0.573 | 0.537 |
| Freight/Risk Model | 0.391 | 0.370 |
| Macro Model | 0.249 | 0.232 |
| Supply Model | 0.157 | 0.138 |
| Variable | Coeff. | p-value |
|---|---|---|
| USD Index | +1.774 | 0.0003 |
| Industrial Production | +1.615 | 0.0024 |
| Baltic Dry Index | +0.005 | 0.0082 |
| Geopolitical Risk | +0.210 | < 0.0001 |
| Dom. Crude Production | −0.009 | 0.0003 |
| CBOE Volatility (VIX) | −0.461 | 0.0439 |
| Inventory Change | −0.0002 | 0.0792 |
| Variable | Coeff. | p-value |
|---|---|---|
| Industrial Production | +12.355 | 0.0005 |
| Geopolitical Risk | +1.360 | < 0.0001 |
| Baltic Dry Index | +0.022 | 0.0544 |
| USD Index | +6.769 | 0.0331 |
| Dom. Crude Production | −0.052 | 0.0009 |
| CBOE Volatility (VIX) | −2.560 | 0.0907 |
| Inventory Change | −0.0009 | 0.1304 |
Demand & risk dominate: Industrial production and geopolitical risk are consistently the strongest positive drivers across both markets.
Supply stabilizes: Inventory change consistently reduces prices — acting as a balancing mechanism between supply and demand.
Structural difference: Brent reflects macro-financial conditions (USD, VIX), while VLSFO behaves more like a logistics-sensitive fuel market driven by physical demand and freight.
Simulating how crude prices respond to changes in demand, freight activity, and geopolitical risk. All impacts represent average price change from the baseline.
| Scenario | Shock Applied | Brent Avg Impact | VLSFO Avg Impact |
|---|---|---|---|
| Geopolitical Shock | +30% GeoRisk | +$7.19 | +$46.48 |
| Demand Boom | +15% IndProd | +$24.30 | +$185.84 |
| Demand Crash | −15% IndProd | −$24.30 | −$185.84 |
| Freight Shock | +25% BDI | +$1.92 | +$9.18 |
| Combined Crisis | +30% GeoRisk, +25% BDI, −15% IndProd | −$15.19 | −$130.18 |
| Demand-Freight Mismatch ⭐ Largest Impact | +15% IndProd, +25% BDI | +$26.22 | +$195.02 |
Industrial production changed ±15%, representing economic expansion/contraction.
Interpretation: Demand shocks produce the largest impact. Economic activity is the dominant driver of crude pricing. VLSFO shows significantly higher sensitivity.
Simulates increased global uncertainty from conflict or supply disruptions.
Interpretation: Moderate impact compared to demand shocks. VLSFO shows stronger response, indicating higher sensitivity to disruptions in global trade and fuel supply chains.
Models an increase in shipping costs representing tighter freight capacity.
Interpretation: Smallest individual impact, but highly relevant from a supply chain perspective. VLSFO again shows stronger response, highlighting its direct linkage to shipping activity.
Strong demand coincides with constrained freight capacity — port congestion scenario.
Interpretation: Highest price increase across all scenarios. When strong demand coincides with limited logistics capacity, price pressures are significantly amplified.
Demand dominates. Even in the combined crisis (geo + freight + demand crash), the negative demand shock overrides all upward pressures.
VLSFO is more volatile. Across all scenarios, VLSFO exhibits significantly larger responses — up to 7× the Brent impact in demand scenarios.
Interaction amplifies risk. The demand-freight mismatch (highest impact scenario) highlights how driver interactions can amplify price movements beyond individual shocks.
This study demonstrates that crude pricing is driven by a combination of demand, supply, and risk factors, with demand playing the most dominant role.
Time-series models confirm price movements are primarily trend-driven. Holt Trend for Brent and Moving Average for VLSFO provide the best short-term accuracy. Seasonal models consistently underperform.
Brent reflects broader macro-financial conditions (USD Index, VIX). VLSFO is more sensitive to physical market dynamics — demand and freight. These structural differences matter for hedging and procurement strategy.
Scenario analysis reveals that demand-freight mismatches generate the greatest price pressure. Monitoring industrial production and geopolitical risk indices provides early warning signals for supply chain cost management.