Deploying integrated predictive analytics raised forecasting accuracy from 71% to 90%, cut disruption probability by roughly 45% and trimmed inflation pass-through by 1.9 percentage points in a year-long institutional trial; however, the evidence comes from a small, nonrandomized sample at a single institution.

Main Finding

Information-system–based predictive analytics applied to 42 industry-linked U.S. supply‑chain datasets over 12 months (Jan–Dec 2024) substantially improved operational forecasting and stability and reduced estimated inflationary pass‑through. Forecasting accuracy rose from 71.4% to 89.7% (p < 0.001); disruption probability fell from 0.42 to 0.23 (p < 0.001); and inflation pass‑through declined by 1.9 percentage points (4.6% → 2.7%, p = 0.012).

Key Points

• Study design: prospective analytical evaluation conducted at Lamar University using 42 complete datasets from manufacturing (n=12), logistics/transport (n=10), retail distribution (n=8), procurement (n=6), and warehousing (n=6).
• Methods: integrated ERP/WMS/TMS/IoT data; supervised ML (random forest, gradient boosting), anomaly detection, multivariate regression; analyses in Python and SPSS; 10‑fold cross‑validation; paired t‑tests, correlations, RMSE/MAE/R² reported.
• Main quantitative results:
  • Forecasting accuracy: 71.4% ± 6.2 → 89.7% ± 4.8 (Δ +18.3 pp, p < 0.001)
  • Forecasting error: 28.6% → 10.3% (Δ –18.3 pp, p < 0.001)
  • RMSE: 14.8 ± 3.1 → 8.6 ± 2.4 (p = 0.002); MAE: 9.3 → 4.8 (p = 0.001)
  • Lead‑time variability: –28.6% (42.6h → 30.4h; p = 0.004)
  • Inventory turnover: +22.4% (4.1 → 5.2; p = 0.009)
  • Supplier delays/month: 6.4 → 3.7 (–42.2%; p = 0.003)
  • Stockout frequency: 17.8% → 8.9% (–50%; p < 0.001)
  • Demand–supply deviation: 14.8% → 6.1% (p < 0.001)
  • Anomaly alerts/month halved (22 → 11; p = 0.001)
  • Inflation pass‑through: 4.6% → 2.7% (Δ –1.9 pp; p = 0.012)
• Associations: disruption probability strongly correlated with inflation pass‑through (r = 0.72, p < 0.001); forecasting accuracy inversely correlated with lead‑time variability (r = –0.63).
• Ethics: IRB approval LU‑ISR‑2024‑1176; anonymized data and data‑use agreements reported.

Data & Methods

• Data sources: timestamped system logs from participating firms’ ERP, WMS, TMS, GPS/IoT telemetry; supplemental macro indicators (e.g., PPI, transport cost indices).
• Sample: 42 datasets selected for completeness; datasets with missing or duplicated essential fields excluded.
• Preprocessing: deduplication, normalization, timestamp sync, IQR‑based outlier flagging, anonymization.
• Modeling pipeline:
  • Supervised ML (random forests, gradient boosting), anomaly detection algorithms, multivariate linear regression for macro linking.
  • Performance metrics: accuracy, RMSE, MAE, adjusted R²; inference via paired t‑tests (α = 0.05) and Pearson correlations.
  • Validation: 10‑fold cross‑validation and internal consistency/manual reconciliation.
• Limitations reported or implied:
  • Sample drawn from voluntary industry partners (42 datasets) — potential selection bias and limited representativeness across all U.S. firms, especially SMEs.
  • One‑year horizon and pre/post comparisons within the same period may confound secular macro trends and causal attribution.
  • Inflation is multi‑causal (monetary policy, labor, commodities); study focuses on supply‑side contributors and uses an “inflation pass‑through” index rather than economy‑wide inflation measures.

Implications for AI Economics

• Micro → Macro transmission: The paper provides empirical evidence that improved ML forecasting and anomaly detection within supply chains can materially reduce operational volatility (lead times, stockouts) and lower estimated inflationary pass‑through from logistics disruptions. If scaled, similar interventions could dampen some supply‑side shocks that feed into headline inflation.
• Policy relevance: Results support policy measures that encourage data sharing, investments in interoperable information systems (ERP/WMS/TMS/IoT), and public–private analytics collaboration to strengthen national supply‑chain resilience and moderate inflationary spillovers.
• Business implications: Firms that adopt integrated predictive analytics can expect lower inventory costs, fewer stockouts, and reduced procurement premiums — improving competitiveness and reducing price pressure on consumers.
• Cautions for scaling and inference:
  • External validity: Effects observed in partner firms may overstate achievable national effects; more representative, larger‑scale evaluations are needed.
  • Attribution: Reduction in inflation pass‑through is plausible but modest (1.9 pp); macroeconomic outcomes depend on broader factors (monetary policy, labor markets, global commodity prices). Careful identification strategies (instrumental variables, difference‑in‑differences across adopters) are needed to strengthen causal claims.
  • Distributional and equity concerns: SMEs often lack the data infrastructure and capital to implement these systems; targeted support may be required to avoid widening resilience gaps.
  • Algorithmic risks: Data quality, interoperability, cybersecurity, and model biases remain practical constraints—governance and standards are important complements to technical deployment.
• Research agenda suggestions:
  • Larger, quasi‑experimental studies linking firm‑level adoption to regional or sectoral price outcomes.
  • Cost–benefit and adoption‑barrier analyses for SMEs.
  • Integration of supply‑chain predictive models with macroeconomic forecasting frameworks to quantify system‑wide inflationary effects under counterfactual shocks.

Summary: The study presents promising evidence that information‑system–based predictive analytics can reduce supply‑chain disruptions and modestly lower inflation pass‑through in participating firms. Scaling these benefits to the national level requires broader uptake, rigorous causal evaluation, and policies to address data, governance, and equity constraints.