Data Analytics Portfolio: Executive Dashboard

Consolidated results across credit risk modeling, multi-asset portfolio risk, and finance operations analytics.
Analyst: ANKIT-INSIGHTS
Reporting period: FY2025 (synthetic, portfolio demo)
Full notebooks & SQL in repo root
Best model ROC-AUC (credit risk)
N/A
Gradient Boosting, beats score-only cutoff
Underwriting policy profit lift
+$4.33M
vs. current 620-score cutoff, test sample
Max-Sharpe portfolio 95% 1-day VaR
$254K
on $10M notional, position-capped allocation
YTD net income (finance ops)
$825K
Revenue $1.48M vs. expenses $656K
Credit Risk Model
Portfolio Risk
Finance Operations

ROC Curve: Model vs. Current Rule

Gradient Boosting and Random Forest separate good/bad loans far better than the credit-score-only cutoff.

Findings

  • DTI, utilization and loan-to-income outrank raw credit score as standalone default predictors (hypothesis-tested, p<0.001).
  • Home ownership & loan purpose are not significant (p=0.92, p=0.43), so they are correctly excluded from the scorecard.
  • Re-optimizing the approval threshold to ≈0.11 default-probability, instead of the legacy score cutoff, is projected to swing net portfolio profit from -$4.19M to +$0.14M on the test sample.

Net Profit vs. Approval Threshold

Expected portfolio profit swept across default-probability approval thresholds.

Top Feature Importances

Gradient Boosting model, relative importance.

Efficient Frontier

8,000 simulated portfolios; Max Sharpe solved with a 35% single-asset position cap.

Findings

  • Unconstrained mean-variance optimization put 100% into Bitcoin. That is a textbook Markowitz instability, not a real recommendation. A 35% position cap produces a deployable allocation.
  • The capped Max-Sharpe portfolio still beats naive equal-weighting on risk-adjusted return.
  • Correlation-shock stress test: 95% 1-day VaR rises ~14% when correlations move toward a crisis regime. Base-case VaR understates true tail risk.
  • EWMA (λ=0.94) volatility forecasting reacts faster to regime change than a 21-day rolling window, making it useful for position limits.

Portfolio Drawdown

Max-Sharpe (35% cap) portfolio, simulated 3-year path.

VaR Under Stress

95% 1-day Value-at-Risk, $10M notional.

Actual vs. Budget by Category

Nine-month actuals (Jan–Sep 2025) vs. scaled monthly budget, from the linked Excel workbook.

Findings

  • Every expense category tracked within budget in the sample period. The largest favorable variances were in Office Supplies (+17.1%) and Utilities (+16.3%).
  • Payroll ran essentially on-budget (-0.7% variance). It is the largest line item, so even a small percentage variance merits close monitoring.
  • The full workbook (Transactions, Assumptions, Summary, Dashboard tabs) uses live SUMIFS / INDEX-MATCH formulas with no hardcoded results, so it recalculates as new transactions are added.

Monthly Revenue vs. Expenses

Jan–Sep 2025, from Excel Summary tab.
All figures on this dashboard are computed in the accompanying Jupyter notebooks and Excel workbook using synthetically generated but realistically-calibrated data (see each project's README for methodology and limitations). This dashboard is a summary view. See flagship-credit-risk/, portfolio-risk-forecasting/, and excel/Financial_KPI_Dashboard.xlsx for full analysis.