Real freight-network risk costs more than history suggests: +25%, not +3%
Assessing resilience with historical disruption scenarios understates systemic risk by an order of magnitude; generated scenarios push expected annual cost to $5.11M.
Key numbers
- Problem: intermodal freight networks face correlated, multi-mode disruptions, yet resilience assessments often rely on historical or uncorrelated scenarios that understate systemic risk.
- Method: a stress-testing framework couples generative adversarial networks with an intermodal optimisation model, feeding generated scenarios into routing problems and aggregating system costs; the case study covers weather-related disruptions in the Tennessee Valley corridor.
- Results: historical disruptions raise total cost by about 3%, while GAN-generated scenarios raise costs by more than 25%, producing an expected annual cost of $5.11 million.
- Risk concentrates in correlated multi-node failures and critical nodes such as the Port of Knoxville, and the framework ranks resilience investments accordingly.
Method and sample
arXiv paper “Data-Driven Stress Testing of Intermodal Freight Networks Using GAN-Generated Disruption Scenarios” (2026-09-15): GAN-generated compound disruption scenarios drive an intermodal optimisation model to estimate costs, with a case study on weather-related disruptions in the Tennessee Valley corridor.
What it means for brands
A logistics firm's resilience assessment is bounded by the failure scenarios it can imagine. The AI parallel is direct: when AI recommends carriers or suppliers it works from retrievable credentials, coverage, lead-time commitments and case history — the more complete your public facts, the less likely you are dropped during risk screening.
How SourceSight handles it
For logistics projects SourceSight checks whether AI can read your credentials and service-scope pages, covers the industry directories and certifications your market expects, and reports per engine so you can tell “not retrieved” from “retrieved but not written into the answer”.
Limits
The case study covers weather-related disruptions in one corridor and does not transfer to other regions or cargo types; the expected annual cost is a modelled figure, not an observed loss record.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
Cross-category recommendations: +36% add-to-cart on item pages, +15% on cart pages
Recommending general merchandise to grocery shoppers — a milk frother with the milk — has long been hard; LLM retrieval plus real-time cart context moved production A/B metrics.
Industries · 2026-10-07Home services: 22% of homeowners already use AI to find a provider, and 80% of businesses don't know how to be found
The local services funnel is moving upstream: before a homeowner opens a website or calls, AI has already synthesised reviews, reputation and service descriptions into one answer.