Regulatory thresholds turn shipment quantity into a discrete decision, so the ADR 1.1.3.6 exemption that lets you move 333 litres of a transport category 2 substance on an ordinary vehicle disappears at 334. Segregation rules make load building a constrained problem that weight and cube models cannot see, and the infeasibility surfaces at the dock. Routing is a permission question before it is a distance question, because tunnel restriction codes and national bans remove links from the network. The pool of certified drivers and approved tank equipment is much smaller than the general freight pool, which changes how a hazmat lane responds to a tender.
A distribution centre has shipped the same solvent to the same customer every week for three years. Twelve cases, 300 litres, on a standard van, one transport document, no orange plates, no specially certified driver. The planner running a consolidation exercise notices the lane and moves it to fortnightly at 600 litres, which halves the shipment count and looks like a straightforward win on the freight line.
The first fortnightly load is refused at collection. The product is a packing group II flammable liquid, which puts it in ADR transport category 2, and the exemption the lane had been running under caps that category at 333 litres per transport unit. At 600 litres the shipment needs a driver holding an ADR training certificate, orange plate marking, an approved fire extinguisher set, and a carrier who accepts full dangerous goods work. The cost per litre goes up rather than down, and nothing in the consolidation model knew that a threshold existed.
That pattern repeats across every hazardous lane a business runs. Classification changes the feasible set, and it does so in steps.
The exemption you were shipping under is a quantity threshold
ADR 1.1.3.6, the provision most commonly called the thousand point rule, assigns every dangerous substance to a transport category and gives each category a multiplier. Category 1 counts fifty points per kilogram or litre, category 2 counts three, category 3 counts one, and category 4 counts nothing. Sum the points across everything on the vehicle, and if the total stays at or below 1,000 the load escapes most of the ADR requirements. Category 0 substances have no exemption at all.
Work the arithmetic and the practical caps fall out. Twenty units for category 1, 333 for category 2, 1,000 for category 3. Those numbers are the real order quantity constraint on a large share of packaged chemical distribution, and they are absent from every economic order quantity calculation I have seen in a chemicals ERP.
The interesting consequence is that the cost curve for a hazardous lane is not smooth. Below the threshold you pay ordinary freight rates from an ordinary carrier. Above it you pay a hazardous rate, from a smaller carrier pool, with a driver who holds a certificate that takes days of training and periodic refresh. The step is large and it sits at a fixed quantity, so any optimisation that assumes cost per unit falls monotonically with shipment size will walk straight through it.
Mixed loads make this sharper. A vehicle carrying 200 litres of a category 2 product and 400 kilograms of a category 3 product sits at 600 plus 400, which is exactly 1,000 and exactly at the limit. Add one more case of anything in category 3 and the whole vehicle loses the exemption, including the products that were fine on their own. Consolidation across products is where the rule bites hardest, and consolidation across products is precisely what a load builder is designed to do.
Limited quantity and excepted quantity provisions give a second set of thresholds, this time on inner packaging size rather than total load. The same substance packed in one litre bottles inside a compliant outer box can travel under limited quantity marking with a much lighter set of obligations than the same substance in a 200 litre drum. Packaging configuration is therefore a planning variable, and it belongs in the item master as a distinct object rather than as a note on the material.
Segregation makes load building a constrained problem
The IMDG Code sets four segregation levels for sea transport, running from away from, through separated from, to separated by a complete compartment or hold from, and finally separated longitudinally by an intervening complete compartment or hold from. Substances also carry segregation group membership, so acids and alkalis and several other families are kept apart on the basis of the group rather than only the class. ADR 7.5.2 applies a similar logic to mixed loading on road vehicles.
The planning consequence is that two items which fit together perfectly on weight and cube may be prohibited from sharing a container. A load builder that optimises fill rate against dimensional and mass limits will publish loads that cannot legally be built, and the discovery happens at the packing station, which is the most expensive place in the chain to find out.
The fix is unglamorous. Segregation is a published pairwise compatibility matrix over groups and classes, it changes slowly, and it expresses as a constraint in a load building model the way a temperature regime does. What stops most businesses is that the attribute lives in an EHS system while the load builder reads the item master, and nobody has connected the two.
Tank container fleets carry the same problem in another form. Previous cargo restrictions mean a tank that carried one product may not take another without documented cleaning to a specified standard, and some products require dedicated service. A fleet of a hundred tanks with three dedicated sub-fleets is four smaller fleets, and pooling calculations that treat it as one overstate available equipment every time.
Routing is a permission problem before it is a distance problem
ADR assigns tunnel restriction codes from A to E to dangerous goods, and each road tunnel in a participating country is assigned a category. A category E tunnel excludes essentially everything carried above the 1.1.3.6 threshold. Several countries operate additional national restrictions, and a number of cities restrict hazardous transport by time of day or exclude it from central zones entirely.
A routing engine scoring on distance, drive time and toll cost will happily route a load through a tunnel it may not enter. The driver takes another route, so actual transit runs longer than the plan and the difference surfaces as a carrier reliability problem. In mountainous corridors a detour around a restricted tunnel can add several hours, enough to break a delivery window the planning system believes is comfortable.
Rail has its own version. In the United States, carriers moving certain highly hazardous commodities must perform a documented route analysis under 49 CFR 172.820, weighing a specified set of safety and security factors and selecting the practicable route with the lowest overall risk. The route a shipper assumes on a map is therefore subject to a carrier determination the shipper does not control, and transit assumptions built from distance will be optimistic on those lanes.
The certified pool is smaller than your tender assumed
Every element of hazardous road transport requires a qualification that takes time to obtain. Drivers need ADR training certificates. Tank vehicles need type approval. Any business consigning dangerous goods in scope must appoint a dangerous goods safety adviser under ADR 1.8.3. Each of those is a supply of qualified capacity that cannot expand quickly in response to a rate.
That changes the economics of a freight tender on hazardous lanes. On general cargo, an aggressive tender moves volume between many carriers competing for it, and capacity follows price fairly quickly. On a hazardous lane in a given region, the number of carriers with the right equipment and certified drivers can be in single figures, so a tender that wins a lower rate can lose access to the only carrier that reliably covers the lane in December. Freight procurement mechanics are covered elsewhere (Z1), and the point here is that the elasticity assumption underneath them changes when the qualification requirement is binding.
Equipment turn time is the same story on the asset side. A tank container cycle is transit out, discharge, cleaning, repositioning and transit back, and the cleaning step queues at a limited number of approved stations. Take a fleet of forty tanks on a twenty-one day cycle: that supports roughly fifty-seven loads in a thirty day month. Let cleaning station congestion add five days to the cycle and the same forty tanks support forty-six loads. Nineteen percent of the fleet's capacity disappeared without a single tank leaving service, and cleaning station lead time appears on nobody's capacity plan.
The expedite option you may not have
Most inventory policy carries an unstated assumption: if the buffer fails, there is an expensive way to recover. Air freight the shortfall, pay the premium, take the hit on the cost line and keep the customer.
For a large part of the hazardous catalogue that option is restricted or absent. The ICAO Technical Instructions and the IATA Dangerous Goods Regulations built on them forbid many substances on passenger aircraft, permit some only on cargo aircraft, and cap quantities per package well below what an ordinary emergency shipment would need. Packing group I flammable liquids and several classes of oxidiser and organic peroxide fall into that territory. The regulations also move: IATA republishes the Dangerous Goods Regulations annually with effect from 1 January, so an eligibility check done eighteen months ago may no longer hold.
The inventory consequence follows directly. The expected cost of a shortage on an item with no viable expedite path is higher than the cost of a shortage on an item that can be flown in, because the recovery time is a full replenishment cycle rather than three days. Sizing both items to the same service target treats them as equivalent when they are not. Safety stock mechanics belong to another post (I2), and what belongs here is the input: the expedite availability flag, per item, per destination region.
What the item master has to carry
The field most chemical item masters hold is a hazardous flag with two values, which triggers a warning and supports no planning decision at all. A usable record carries the UN number, the proper shipping name, the class and any subsidiary risk, the packing group, the transport category and its multiplier, the tunnel restriction code, IMDG segregation group membership, marine pollutant status, the applicable portable tank instruction where relevant, and limited quantity eligibility for each packaging configuration sold.
That last clause is the one that gets missed. These attributes attach to a substance and packaging combination rather than to a material number. The same product in five litre bottles, in a 200 litre drum, and in a 20,000 litre tank container is three transport objects with different thresholds, different marking, and in some cases different modal eligibility. An item master keyed only on material will collapse them and produce answers that are right for one configuration and wrong for the others.
The data usually already exists, in a safety data sheet management system maintained by regulatory affairs. Getting it into the planning model is an integration exercise rather than a data creation one, which makes it cheaper than it looks and argues for reading it live rather than holding a copy.
Where this stops
Classification is a regulatory determination made by people who carry legal responsibility for it. A planning system consumes a classification, applies its consequences, and never derives one. Any model that infers a UN number from a formulation is producing a guess with liability attached, and the correct behaviour when an attribute is missing is to refuse to plan the shipment rather than to assume the benign case.
The regulatory attributes also expire. ADR is revised on a two year cycle with a transition window, the IMDG Code follows its own two year amendment cycle, and IATA changes annually. A cached transport category or tunnel code is accurate until it is not, and the planning model should carry the source version alongside the value so that a revision triggers a review rather than passing unnoticed.
There is a further limit on coverage. National derogations, port specific rules, municipal restrictions and individual carrier policies sit below the international instruments and are not available in any single dataset. A model carrying the global attributes correctly will still be wrong about a particular city on a particular day, which means the planning system's job is to eliminate the systematic errors and leave the local ones to the people who know the lane. Expect that to cut refused loads substantially while leaving a residue that only lane knowledge catches.
Take the fifty highest volume hazardous item and packaging combinations in your catalogue and check whether the item master holds the transport category and the tunnel restriction code for each. If what you find is a yes or no flag, you have located the gap, and the check takes an afternoon.
More on this in industry on the DemandLabs platform.
