Supply Chain Analytics Pulse

Why Manufacturers Kept Stockpiling After Their Tariffs Got Struck Down

September 2026 | Supply Chain Analytics Pulse

In February 2026, the Supreme Court struck down the legal authority behind most of the tariffs that had pushed U.S. manufacturers into their biggest inventory buildup in a decade. The tariffs were gone within days. The stockpiling was not.

Four months later, GEP’s Global Supply Chain Volatility Index, which tracks tens of thousands of businesses each month, found manufacturers building safety stock at the highest level since January 2023. Shortages of critical inputs hit their highest level since late 2022.

Meanwhile, U.S. factories just posted their strongest month of production since 2021, and the index tracking how “too low” customers’ inventories are sits near a multi-year extreme. A replenishment cycle is forming at exactly the moment safety stock strategy needs to get smarter, not just bigger.

01 Eighteen Months of Tariff Whiplash

The 2025 Buildup

Through 2025, U.S. manufacturers front-loaded raw materials and components like never before. Durable-goods “materials and supplies” inventories hit roughly $221 billion by April, up 3.9% year over year, and total durable-goods inventories climbed for ten straight months, to about $591 billion by July.

Trade uncertainty drove it. NAM’s Outlook Survey found it was manufacturers’ top business challenge all year, cited by 76.2% in Q1 and 73.1% in Q4. By year’s end, 80.3% had paid tariffs on imported inputs, and the Hackett Group put a price on the caution: $1.7 trillion trapped in excess working capital across the 1,000 largest U.S. public companies.

Through 2025, U.S. manufacturers front-loaded raw materials and components like never before. Durable-goods “materials and supplies” inventories hit roughly $221 billion by April, up 3.9% year over year, and total durable-goods inventories climbed for ten straight months, to about $591 billion by July.

NAM respondents expected the buildup to ease slightly, about 0.2%, over the next year. The stockpiling wave, it seemed, had peaked.

The reversal nobody de-stocked for

On February 20, 2026, the Supreme Court ruled 6-3 in Learning Resources, Inc. v. Trump that IEEPA does not authorize the president to impose tariffs, striking down the “Liberation Day” and reciprocal tariffs behind most of 2025’s front-loading. Collection stopped four days later. Refunds are still working through the Court of International Trade.

What replaced IEEPA moved just as fast:

Section 232 tariffs on steel, aluminum, and copper run on separate legal footing and were untouched by the ruling; they remain at 50% on covered articles. What changed, effective April 6, 2026, was scope: many derivative products, parts and components made from those metals, moved to a flat 25% duty on full import value instead of just the metal content, quietly widening exposure for metal-intensive manufacturers.

The data did not cooperate with the “hangover” prediction

Here’s what the original 2025 forecast expected: demand cools, shelves stay full, and 2026 becomes a quiet year of working through last year’s buying. That’s not what happened.

Materials-and-supplies inventories reached $225.5 billion in January 2026, still above where they stood in April 2025. Total durable-goods inventories kept climbing too, up nine straight months to $602.0 billion by June. GEP’s June figures told the same story from the buying side: safety stockpiling back at its highest level since January 2023.

Demand didn’t cooperate either. The ISM Manufacturing PMI hit 55.6% in July, the best reading in four years, while the index tracking customers’ inventories sank to 40.7, deep in “too low” territory.

Put those two trends together and the picture isn’t a hangover. It’s a restocking cycle arriving on top of a warehouse that never actually emptied.

02 What “Segment and Go Dynamic” Actually Delivers: A $9.3M Proof Point

Segment inventory first and make safety stock dynamic rather than static, is not a theoretical improvement. MIT’s Center for Transportation and Logistics ran the numbers on a real network, and they are large enough to change how a safety stock budget gets built.

The sponsor was a U.S. grocery retailer running a hub-and-spoke distribution network that was, by its own admission, missing the mark despite carrying plenty of stock: 57 days of supply for dry food, low inventory turnover, and service levels that still varied widely from node to node. As the business had expanded and SKU counts multiplied, a traditional inventory policy, one set of rules applied network-wide, had stopped working.

Researchers Vi Duong and Nic Holwerda, supervised by Dr. Eva Ponce, modeled 61 SKUs across 31 nodes using a commercial multi-echelon inventory-optimization platform, running 18 scenarios that combined six update frequencies with three service-level targets. Published through Supply Chain Management Review in March 2026, it’s one of the more concrete proof points available for what dynamic, segmented safety stock actually returns:

High-variability products benefited the most from frequent updates. Stable, low-variability SKUs saw little additional gain from updating more often than biannually: the highest-value move wasn’t updating everything weekly, it was segmenting products by variability and matching update frequency to each segment.

That last point maps directly onto how tariff-exposed inventory should be triaged. The SKUs and inputs worth a dynamic, frequently recalculated safety stock policy are not the highest-volume ones by default. They are the ones with genuine demand or supply volatility, which, in 2026, increasingly means anything touched by a tariff schedule with a defined expiration date, a single-sourced input, or a lane running through a contested shipping corridor.

03 Two Sectors, Two Strategies: Automotive’s Live Split Test

Detroit’s three largest automakers spent the first half of 2026 absorbing the same tariff and input-cost environment and arrived at two different buffer strategies, in real time, one of the clearest live comparisons available for how the same volatility produces different inventory decisions depending on where a company’s actual exposure sits.

04 Safety Stock Readiness Check

Before adding a single unit of buffer inventory, it’s worth an honest look at whether the organization can actually target that buffer, or whether it’s about to repeat the blanket-stockpiling pattern that trapped $1.7 trillion in working capital industrywide. 

05 Key Implementation Steps: Choose Your Perspective

Click on the perspective you want to analyze – executives or analysts.

06 Conclusion

The last eighteen months did not validate or discredit the manufacturers who built safety stock through 2025. It discredited the idea that safety stock built to hedge one specific policy risk will still make sense once that policy changes, because in 2026, policy changed faster than most inventory plans do.

What held up across that whiplash was not a bigger buffer. It was a better-targeted one. Buffer inventory earns its cost when it’s aimed at something genuinely concentrated, volatile, and hard to substitute. It’s just tied-up cash everywhere else.

The restocking cycle now forming, with customer inventories near multi-year lows and factory output at a four-year high, is the next test of that discipline. It will reward the organizations that already know which SKUs deserve the buffer, and it will quietly cost the ones still deciding that with a blanket policy and a calendar reminder.

Supply Chain Analytics Pulse

The New Cost of Picking the Wrong Carrier

August 2026 | Supply Chain Analytics Pulse

In June 2026, the average U.S. dry van spot rate did something it hadn’t done since February 2022: it rose above the contract rate. Tender rejections hit 17.55%, the highest level in four years.

Three weeks earlier, the U.S. Supreme Court had ruled  that freight brokers can be sued under state law for carelessly choosing an unsafe carrier. And across the industry, fraud investigators were closing out 2025 with cargo-crime losses near $725 million, up 60% year over year.

None of these three developments made headlines together. But put next to each other, they describe the same moment from three different angles: the decision of which carrier to use just got harder to make well, and more expensive to make badly.

01 Three Forces, One Decision

Capacity is turning for the first time in years

From 2022 through most of 2025, trucking capacity ran well ahead of freight demand. Carrier counts had ballooned during the pandemic-era boom, and the resulting oversupply pushed rates down for nearly four straight years, the longest freight recession most shippers had ever managed through. That gave buyers enormous leverage: if one carrier’s price or terms weren’t ideal, ten more were waiting.

That leverage is disappearing. According to DAT Freight & Analytics, the national average dry van spot rate crossed above the contract rate in June 2026 for the first time since February 2022, with spot rates briefly touching $3.83 per mile. Ryder and FreightWaves’ June State of Transportation report put tender rejection rates above 15%. Reefer and flatbed segments show the same pattern.

The mechanics matter here: when carriers can earn more chasing spot loads than honoring contracted rates, they reject contracted tenders. Shippers scramble to fill the gap, often with whichever carrier is available on short notice.

The legal shield brokers relied on just narrowed

On May 14, 2026, the U.S. Supreme Court issued a unanimous decision in Montgomery v. Caribe Transport II, LLC, holding that a federal transportation law does not automatically shield brokers from state negligent-selection lawsuits when a carrier they chose turns out to be unsafe. To be precise about what that does and doesn’t mean: the Court didn’t rule on anyone’s guilt, it removed a procedural defense that used to get these claims dismissed early, before discovery. The practical effect is that negligent-selection claims against brokers are now harder to dismiss nationwide, and legal commentary suggests the same reasoning could plausibly extend to other intermediaries that select carriers, including 3PLs and digital freight-matching platforms, though that hasn’t yet been tested in court.

Fraud is rising exactly when vetting is hardest to do carefully

The third force is less visible in boardrooms but well documented in industry data. According to Verisk CargoNet, U.S. supply chain crime losses reached roughly $725 million in 2025, up 60% from 2024, with the average loss per incident climbing to $273,990. CargoNet separately reports that “strategic theft”, fraud and identity-based schemes, as opposed to physical break-ins has risen roughly 1,500% since 2021.

Licensed freight broker based in San Antonio, had its identity stolen by overseas criminals who spoofed its email domain and logo, then brokered real loads to unsuspecting carriers under their name. One six-figure shipment of energy drinks was rerouted more than a thousand miles before anyone realized the company that had booked it wasn’t real. The fraudulent entity was even added to the FMCSA’s public SAFER safety-rating database alongside the legitimate one, and owner Adam Blanchard testified that getting it removed proved difficult.

02 What the Research Shows

On detection performance

May 2026 study by Hayes and colleagues, published via a machine learning and behavioral analytics framework for supply chain fraud, combined graph neural networks with behavioral pattern analysis and reported 0.94 precision and 0.91 recall in identifying fraudulent transactions, a 34% reduction in false positives compared to standalone models. Notably, the behavioral component surfaced collusive fraud rings involving shell vendors and inflated invoicing that traditional rule-based checks had missed entirely. Pattern-based models that update continuously can catch coordinated fraud that this kind of one-time check will never see, simply because they keep watching after the initial check is done.

On legal treatment of automated decisions

The law is moving faster than many logistics organizations may have registered. California’s AB 316, effective January 1, 2026, prohibits a company from defending itself by arguing that an AI system “autonomously” caused the harm: a company that deploys, modifies, or uses an AI tool cannot point to the tool as the responsible party. A June 2026 analysis in the Berkeley Technology Law Journal goes further, arguing that liability frameworks built around a single AI system don’t hold up well once multiple AI agents interact, for instance, a shipper’s procurement agent negotiating with a carrier’s booking agent, because those handoffs are typically unlogged and difficult to reconstruct after something goes wrong. NIST’s AI Agent Standards Initiative, launched in February 2026, has identified agent identity and traceability as a priority precisely because of this gap. A useful, non-freight precedent is Mobley v. Workday, in which a plaintiff argued that an employer couldn’t avoid discrimination liability simply because an algorithm made the screening decision; courts have shown little appetite for treating “the algorithm decided” as a standalone defense.

03 How the Industry is Responding

Two proof points and one cautionary tale — chosen because they span three different sub-tier verification methods and three different outcomes.

04 Organizational Readiness Check

If most of your answers sit in Levels 1–2, the current market conditions (tight capacity, active fraud, sharpened legal exposure) describe real risk sitting inside a normal-looking process, not a hypothetical one.

05 What to do Next: Choose Your Perspective

Click on the perspective you want to analyze – executives or analysts.

06 Conclusion

None of these three forces created the others. But they’re now converging on a single, frequently repeated decision that most organizations have treated as low-stakes for a long time: which carrier gets the load. The tools to make that decision more rigorously already exist.

Supply Chain Analytics Pulse

Most Supply Chains Still Go Dark Below Tier 1

August 2026 | Supply Chain Analytics Pulse

A decade-old 6% statistic still gets recycled as current research. The real numbers are different, and the gap between what executives believe and what they can actually verify is the more useful numbers for a board conversation.

The claim that only 6% of companies have full end-to-end supply chain visibility traces back to the GEODIS Supply Chain Worldwide Survey, a poll of 623 companies conducted in 2017. Nearly a decade later, that same 6% figure is still being presented as fresh 2026 data across dozens of procurement and logistics roundups.

More recent, purpose-built surveys put genuine end-to-end visibility somewhere between 13% and 18%. QIMA’s 2026 Global Sourcing Survey found 18% of companies report full end-to-end visibility, an improvement on prior years, though the same survey found average supplier-network mapping sits at only 60%, meaning a large share of companies that have mapped their network still don’t have verified visibility into it. Achilles’ Global Supplier Risk and Sustainability Survey found only 6% of organizations have full visibility specifically into Tier 2 and Tier 3 suppliers, and EcoVadis’s Sustainable Procurement Barometer found only 12% can monitor more than half of their Tier 2 base at all.

The pattern holds regardless of who ran the survey or how “full visibility” was defined: visibility is high at Tier 1 and collapses sharply beyond it. McKinsey’s 2025 supply chain risk research found 95% of leaders report visibility into Tier 1 risk, while only 42% of the same group report visibility into Tier 2 or beyond. Executive confidence in overall visibility runs as high as 93% in some surveys, even as the same executives name Tier 2 and Tier 3 suppliers as their single biggest blind spot. That confidence-capability gap, not any single percentage, is the more useful number for a board conversation: dashboards built on Tier 1 completeness are routinely mistaken for full-chain visibility.

01 Why Technology Alone Hasn’t Closed the Gap

Five structural forces explain why more AI and more dashboards haven’t moved the needle much on their own.

02 Three Methods Actually Closing the Sub-Tier Gap

None of these approaches are new in concept, but each has moved from pilot to production use over the past two to three years in ways that don’t show up in a typical visibility software comparison article.

Trade-data network inference

Rather than waiting for a Tier 1 supplier to voluntarily disclose its own upstream relationships, this approach reconstructs the supplier network directly from customs and bill-of-lading records, information that governments already require at the border regardless of whether any individual company wants to share it. Platforms built on this model, S&P Global’s Panjiva being the largest, now index billions of shipment records and millions of companies across roughly twenty countries, and use network analysis to surface buyer-supplier relationships that would otherwise stay invisible. The method has a real limit worth naming: it shows trade flows and corporate relationships, not labor conditions or compliance status. It is a mapping layer, useful for finding out who is actually connected to whom, not a verification layer that confirms a supplier is behaving responsibly.

Forensic and isotopic origin verification

This addresses a different failure mode: paper trails that are technically complete but unreliable, whether through honest error, transshipment, blending, or deliberate falsification. The method analyzes naturally occurring chemical and isotopic variations in a raw material itself, shaped by the soil, water, and climate where it grew or was mined, and compares that chemical fingerprint against a reference database of known origins. Because the test examines the physical material rather than accompanying documents, it produces evidence that doesn’t depend on a supplier’s paperwork being honest. US Customs and Border Protection has explicitly recognized isotopic testing as admissible evidence for forced-labor compliance and began expanding its own in-house testing capability for exactly this reason.

Blockchain-based material passports

This method tackles the standardization and incentive problems directly, giving multiple parties in a value chain — miner, refiner, component maker, assembler — a shared, tamper-evident ledger to record material provenance, emissions, and recycled content as materials change hands. Rather than asking each sub-tier supplier to disclose data into a buyer’s proprietary system with no benefit to the supplier, a shared passport standard gives every participant in the consortium the same verified record, which is part of why these programs have found more sub-tier cooperation than bilateral disclosure requests typically do. The EU’s Digital Product Passport framework, starting with batteries, is effectively forcing this approach into a common data schema across the industry rather than leaving it to each company to build its own.

03 Case Studies

Two proof points and one cautionary tale — chosen because they span three different sub-tier verification methods and three different outcomes.

04 Organizational Readiness Check

Before committing further budget to visibility tools, it’s worth assessing where the organization actually sits across four dimensions, not just how much has already been spent.

05 Key Implementation Steps: Choose Your Perspective

Click on the perspective you want to analyze – executives or analysts.

06 Conclusion

The visibility gap below Tier 1 has stopped being a technology-availability problem, if it ever really was one. Every method described in this edition — trade-data inference, forensic verification, blockchain material passports — already exists and is already running in production at real companies today. What’s different about the organizations actually closing the gap is that they paired the technology with a contractual mechanism to reach past Tier 1, a verification method that doesn’t depend on a supplier’s paperwork being honest, and a roadmap sequenced to the regulatory dates that are now forcing the issue whether a company is ready or not.