In July 2026 the European Commission's transport directorate recorded that around 95% of Ukraine's non-agricultural exports left the country through the EU-Ukraine Solidarity Lanes, against roughly 5% by sea. Taken as evidence that the land corridors have matured into a dependable route, the figure misleads. Three months earlier the same series put the split at close to half and half. The July reading describes a maritime failure rather than a land-corridor capability, and a buyer who built a continuity plan around it would have reached the wrong conclusion.
This is the recurring analytical error in Ukraine supply chain risk. Exposure is treated as a single country rating, or as one question about whether goods can still leave. Neither framing survives contact with a real supply chain, because continuity depends on the weakest material dependency in one specific flow, and that dependency is seldom the one dominating the reporting.
Ukraine operates several logistics systems, not one
The Commission's July 2026 figures showed grain, oilseeds and related products moving the other way, roughly 20% overland and 80% through Black Sea ports. Both readings are monthly tonnage shares derived from Ukrainian customs registers, and both move sharply. In April 2026 the non-agricultural split was close to even and grain ran at about 10% overland. Measured cumulatively since May 2022, of the roughly 255 million tonnes of grain and oilseeds exported, the Lanes carried 40% and the ports 60%.
The divergence between cargo types has economic causes. Bulk agricultural volume is maritime because no land mode approaches the cost per tonne of a deep-water vessel, and because the loading infrastructure sits at the berth. Industrial cargo behaves differently. Components and processed materials carry enough value per tonne to absorb rail or road economics, which is why buyers of Ukrainian industrial goods retained a genuine land option throughout. A company sourcing pressed steel parts and a company buying sunflower meal from the same country therefore face different bottlenecks and different failure modes. Logistics from Ukraine into the EU is not a single system, which is why supply questions in industrials and manufacturing diverge so early from those in agribusiness and food.
What the 2026 port disruption revealed about substitution
An alternative route is only as useful as the volume it can absorb. Ukraine's Ministry for Development of Communities and Territories reported that seaports handled 46 million tonnes in the first seven months of 2026, of which 42.2 million tonnes moved through the Greater Odesa cluster and 3.8 million tonnes through the Danube ports. Greater Odesa carries around 90% of Ukraine's agricultural exports. The Danube cluster, routinely described as the fallback, is roughly a tenth of the size of the system it is expected to replace.
When vessel calls at Greater Odesa fell away from late July 2026, that arithmetic showed up in the substitution rates. The Ministry of Agrarian Policy and Food reported the following volumes on 11 September 2026.
Product | Moved via road, rail and Danube, early September 2026 | Share of the volume normal logistics would carry |
|---|---|---|
Grain | 380,000 tonnes | About 35% |
Oilseeds | 94,000 tonnes | About 60% |
Vegetable oil | 97,000 tonnes | About 66% |
Meal | 60,000 tonnes | About one third |
All agricultural products | 630,000 tonnes | About 40% |
Substitution therefore depends on the product, not on the corridor. Vegetable oil transfers onto rail tank wagons reasonably well. Bulk grain does not, because the binding constraint is loading equipment, wagon availability and terminal throughput rather than track. Low summer water levels on the Danube removed part of the barge option in the same period, with industry estimates in mid-2026 putting Danube freight around 70% above deep-sea rates. Two alternatives can fail for unrelated reasons in the same week.
Failures do not travel in one direction
The standard mental model runs from supplier to transport to border to late delivery. Real disruption starts anywhere and propagates in both directions. Ferrexpo shows the pattern. The company suspended iron ore production in Ukraine in August 2026 after Black Sea shipping became unusable, partly to preserve working capital, and continued supplying European steelmakers from existing stockpiles. Nothing had happened to the mine. A logistics failure travelled upstream, stopped production and was absorbed downstream by inventory rather than by routing.
Rail is the dependency buyers most often underestimate, because it enters the plan as the solution and never as an exposure. Ukrzaliznytsia, a single national operator, moves more than 90% of the country's export shipments. Ukrainian authorities recorded 1,534 attacks on railway infrastructure between January and late August 2026, with 492 locomotives damaged since February 2022 and rolling stock increasingly targeted. Rail is at once the substitute for lost maritime capacity and a system under direct pressure, so the two routes are not independent.
Power belongs in the same category. A generator sized for refrigeration, lighting and process control is not the asset that runs a production line at rated output, and installed capacity rarely equals normal throughput. The regulatory threshold is a useful proxy. Under a government decision of 19 August 2026, industrial and commercial consumers avoid electricity restriction measures only where they cover at least 80% of consumption from own generation, local distributed generation or imports. A supplier below that level remains subject to the outage schedule. Facility-level resilience design belongs with energy security and resilience planning. A buyer's question is narrower, and concerns whether the supplier's arrangement preserves the output it has committed. Workforce constraints operate more quietly and can cut output while transport runs normally, with Ukrainian industrial surveys putting mobilisation as the leading factor limiting staffing.
Failure layer | What interrupts supply | What reveals the real exposure | Continuity response that fits |
|---|---|---|---|
Supplier production | Damage, equipment failure, single line or single site, maintenance backlog | Whether the specified item can be made anywhere else with the same tooling and approvals | Qualified second site, tooling ownership, transferable specification |
Power and workforce | Restriction schedules, fuel supply to generators, mobilisation, skills loss | Output actually achieved during the last restriction period, not installed generator capacity | Verified autonomy level, production prioritisation, output-based commitments |
Critical inputs and sub-suppliers | Imported material, specialised consumables, packaging, single-origin inputs | Whether the supplier's own inventory and input lead times are known or merely assumed | Input mapping, supplier-held buffer, approved substitute materials |
Domestic transport to port or border | Wagon and locomotive availability, damaged junctions, restrictions on specific stations | Whether the cargo can physically reach more than one loading point | Second loading point, road leg for part of the volume |
Border crossing and transhipment | Gauge change, terminal and crane capacity, congestion, inspection throughput | Which single node the whole volume passes through | Second crossing already used at low volume, intermodal option |
Maritime route and port access | Vessel availability, insurance withdrawal, terminal damage, draught limits | Whether the product has any economically viable land route at all | Inventory positioned outside Ukraine, contracted land allocation |
EU-side warehousing and onward flow | Customer-approved storage, handling capacity, onward haulage | Whether the receiving side can absorb an irregular arrival pattern | Second approved warehouse, buffer nearer the buyer |
Locating the constraint that decides delivery
The visible problem is rarely the binding one. For a manufacturer of machined components the limit is frequently a single heat-treatment or coating subcontractor, and no corridor is involved at all. For a food processor it can be a customer-approved cold store. For bulk minerals it is berth access, and nothing on land substitutes for it.
Finding the constraint takes a physical map of the flow, not the commercial chart of who invoices whom. A useful map records each leg and mode, every transfer point, the specific border node or berth used, the warehouses, the named carriers and terminal operators, realistic transit time with its observed variability and the practical alternative at each step. Two features matter more than completeness. The first is any node that appears twice, because that is where diversification is nominal. The second is any alternative that exists on paper but has never carried volume, since an untested route is an assumption rather than a capability.
Two suppliers, one dependency
Dual sourcing protects against supplier-specific failure and very little else. Two Ukrainian suppliers commonly share a sub-supplier, an imported input, a rail crossing, a terminal, a specialist carrier, a customer-approved warehouse or the same certification body. If both depend on the same berth, one maritime interruption removes both. Supplier redundancy is cheap to arrange and easy to present internally. Dependency redundancy means identifying the shared nodes and deliberately separating at least one of them, which takes longer and usually costs more.
The same test applies to routing. A second corridor counts as continuity only if the product can physically use that mode, if capacity is available at the volume required and not merely in principle, if the transit time change is tolerable and if the receiving customer accepts the resulting handling and documentation. Where those conditions fail, the alternative is theoretical.
Continuity measure | Failure it mitigates | Where protection stops | Principal cost or new dependency |
|---|---|---|---|
Parallel source outside Ukraine | Country-level, supplier-level and corridor-level interruption | Only holds where the second source is genuinely qualified and kept in production | Qualification and audit cost, volume split raises unit price, engineering time |
Buffer stock | Short interruptions of known duration, and lead-time variability | Fails the moment the interruption outlasts the cover held | Working capital, storage, obsolescence and shelf-life exposure |
Reserved capacity or shipment priority | Competition for scarce output or space when several customers are hit together | Worth nothing without demonstrated ability to allocate under stress | Premium or committed volume, weaker position on price |
Transferable production | Loss of one manufacturing site | Blocked by tooling, technical files, approvals, certification and ramp-up time | Duplicate tooling, requalification, customer re-approval |
Alternative routing | Failure of one node, corridor or mode | No protection where routes share an upstream node or the cargo cannot switch mode | Higher freight rate, longer transit, additional handling |
Alternative inventory location | Interruption between the supplier and the buyer's region | No protection if the stock still sits inside the disrupted geography | Warehouse cost, duplicated handling, reduced flexibility |
Inventory and priority are sized against an assumed duration
Buffer stock is the measure most often chosen and least often sized. The governing variable is the length of interruption the company intends to absorb, since four weeks of cover against a twelve-week disruption moves the date of the problem and not the outcome. That estimate then interacts with replenishment lead time and its variability, product value, carrying cost, shelf life and the cost of downtime at the buyer's own plant. Where downtime is expensive and the component is cheap, generous cover is easy to defend. Where the item is high-value or perishable, the honest answer is usually a smaller buffer combined with a second source.
Location changes what a buffer protects against. Stock at the supplier is cheapest and covers only production interruption. Stock elsewhere in Ukraine stays exposed to the same corridor and border risk, and stock near the border still depends on one crossing. Stock inside the EU or at the buyer's site is the only position genuinely insulated from a Ukrainian interruption, and the most expensive in working capital. Reserved capacity deserves similar scepticism. A clause granting priority is not evidence of capability, and the question worth asking is how the supplier allocated output the last time several customers were affected at once.
Testing the plan and documenting what remains
A continuity document is not evidence that continuity will work. Credibility comes from specificity. A plan worth relying on names its failure scenarios individually, identifies alternative production arrangements with stated technical compatibility, reports autonomy measured during actual outages, maps critical inputs with real lead times, states actual inventory policy and commits to a recovery-time assumption the supplier is willing to defend. Evidence of how that supplier behaved during a previous interruption outweighs all of it.
The commercial operating layer needs the same discipline. Notification triggers, escalation routes, access to continuity information, production and shipment priority, inventory visibility and recovery reporting all have to be agreed before they are needed. Contractual drafting, enforceability and the interpretation of force majeure should be settled with qualified legal counsel.
Customers and procurement committees increasingly ask for supply-chain transparency, and generic reassurance fails that test. A defensible representation separates exposure, meaning which products and volumes depend on which Ukrainian nodes, from mitigation, meaning what is in place and has been tested, from residual risk, meaning the interruption length that would still cause failure. The contingency option and its activation time complete the picture. This is a different exercise from planning the continuity of a company's own Ukrainian operation, which is a crisis management questionand not a buyer-side one.
What has improved structurally, and what has not
Part of the change since 2022 is permanent. Ukraine and Moldova joined the trans-European transport network in 2024, and the EU has mobilised more than EUR 2.3 billion for the Solidarity Lanes, including EUR 1.55 billion in Connecting Europe Facility grants across 38 transport projects. Concrete deliveries include the relocated train inspection point at Mostyska II, opened in January 2025, and Ukraine's first standard-gauge railway line, the 22 kilometres between Chop and Uzhhorod, opened in September 2025. The road transport agreement with the EU, which removes permit requirements for bilateral and transit haulage, was extended in September 2025 and now runs to at least March 2027.
The distinction between delivered and planned matters here. The standard-gauge network does not yet carry freight beyond Uzhhorod towards Lviv or Kovel, so the gauge change at most crossings remains a physical requirement through the rest of this decade. Infrastructure spending also does not alter the underlying geography. Bulk commodities still need deep water, the western crossings still form a narrow set of nodes, and rail remains one national operator under attack. Continuity design for Ukrainian supply should therefore assume that the corridor mix will keep shifting and that the decisive constraint will keep moving between production, power, inputs and specific transfer nodes. Designing for that is execution work, and it is where operations consulting and execution support earns its place, converting a dependency map into supplier interfaces, allocation agreements, inventory positions and monitoring that catches a constraint moving before a delivery is missed.
