Premium freight gets reported as a transport number and it is almost never a transport problem. The single most useful statistic in this whole area is that roughly 49% of expedite events trace back to inaccurate demand forecasts — nearly half of all emergency orders are a planning failure that arrives dressed as a logistics cost. Which means the department that receives the invoice is rarely the department that caused it, and the department that caused it never sees the number. That is why premium freight spend stays flat through years of carrier negotiation: the negotiation is aimed at the price of the symptom. Root cause analysis fixes this only if it is honest about attribution, and honest attribution is uncomfortable, because the answer is frequently internal. Book an architecture review to categorise and attribute your own expedite events.
ENTERPRISE GUIDE · PREMIUM FREIGHT
Premium Freight Root Cause Analysis
A cause taxonomy that separates supplier miss from internal planning, attribution rules written before the argument, and the fixes that actually move spend rather than the ones that move blame.
Why did we air-freight this?The part would have missed the line.
Why was it going to miss?The supplier shipped late.
Why did they ship late?The call-off changed inside their lead time.
Why did it change?The build plan was amended after the frozen window.
Most investigations stop at line two, where the cause is someone else's. The cost is attributed to the supplier, the corrective action is a phone call, and the same event recurs next month.
The Benchmark Worth Knowing
Before categorising anything, establish where you sit. The spread between good and poor performance here is wide enough that the gap alone usually funds the programme.
3%
of freight budget spent on expediting by top-performing supply chains
10%
spent by bottom performers — and that gap, multiplied by annual logistics spend, is entirely recoverable
Treat the frequency as the signal
Emergency orders are not a cost of doing business — they are a planning signal, and the frequency of your expedited freight and emergency supplier orders tells you precisely how reactive your supply chain is. A rising count is diagnostic before it is expensive. Report the event count alongside the spend, because spend can be flattened by negotiating rates while the underlying failure rate climbs.
The Cause Taxonomy
Eight categories, each with a different owner and a different fix. The discipline is that every expedite event gets exactly one primary category — not "supply chain".
PLNForecast inaccuracy
Demand was mis-forecast and the requirement appeared inside supplier lead time. The single largest category by a wide margin.
OwnerDemand planning
AMDLate plan amendment
The build plan changed after the frozen window, invalidating a call-off the supplier was already building to.
OwnerProduction planning
SUPSupplier capacity or performance
The supplier could not produce to the committed schedule for reasons within their control.
OwnerSupplier quality and purchasing
QLTQuality rejection
Material arrived and could not be used, so the replacement had to move at speed. Attributable, but not to transport.
OwnerSupplier quality
TRNTransport service failure
The carrier missed. Genuine, and much smaller as a share than most people assume before they measure it.
OwnerTransport
RTERouting or tender failure
Late tender, routing guide exhausted, or spot procurement forced by a gap in the carrier network. Costs accumulate here largely unnoticed.
OwnerTransport procurement
INTInternal handling or yard
Material arrived on time and did not reach the line — gate queue, dock backlog, staging error, misplaced stock.
OwnerPlant logistics
EXTGenuinely external
Weather, port disruption, conflict-driven routing change, regulatory action. Real, and the category most abused as a catch-all.
OwnerNobody — but it still needs recording
Watch the EXT count. Where genuinely external causes account for a large share of your events, one of two things is true: your network is unusually exposed, or the category is absorbing events that belong elsewhere. It is almost always the second, because EXT is the only category that costs nobody anything to select.
Attribution
One primary category per event, chosen by rule rather than by whoever is in the room.
An architecture review sets the taxonomy against your own event history, writes the attribution rules, and shows what the split actually looks like once "supply chain" is no longer an option.
Attribution Rules
Write these before the first contested case, because attribution decided case by case always drifts toward whoever is least present in the meeting.
1One primary cause per event. Contributing factors can be recorded, but the cost lands in exactly one place or the total never reconciles.
2Ask why at least four times. Stopping at "the supplier was late" attributes the cost correctly and identifies nothing.
3A change inside the frozen window is internal, whatever happened downstream of it.
4A requirement appearing inside supplier lead time is a planning cause, not a supplier cause.
5Where the material arrived on time and still missed the line, the cause is internal handling, regardless of who shipped it.
6External requires evidence — a named event, not an unexplained delay.
7Every event names a person, not a function. Departments do not write corrective actions.
8The attribution is recorded at the time, not reconstructed at month end from memory and an invoice.
The reframe that makes the programme work
Link freight cost to the internal decisions that created it — late tenders, poor forecasting, order volatility, emergency shipments — so that premium freight spend reads as a diagnostic of planning discipline rather than as logistics performance. Once the number is presented that way, it reaches the people who can move it. Presented as a transport number, it reaches people who can only negotiate the rate of something that should not be happening.
Fixes That Match the Cause
Each category has a fix that works and several that do not. Carrier renegotiation appears in exactly one row.
← Swipe to see all columns →
| Category | The fix that works | The fix that does not |
|---|---|---|
| Forecast inaccuracy | Forecast accuracy measurement by part family, with the error fed back to planning | Buying more buffer, which hides the error and pays for it twice |
| Late plan amendment | Amendment volume inside the frozen window tracked as a metric with a named owner | Asking suppliers to be more flexible about a window you are not honouring |
| Supplier capacity | Capacity verification at rate, plus reservation agreements on constrained parts | Escalation calls that produce commitment without capability |
| Quality rejection | Detection moved earlier — at pack or at receipt rather than at the line | Increasing incoming inspection sample size after the fact |
| Transport service failure | Tender acceptance and lane execution consistency tracked by carrier | Adding carriers, which increases service variability rather than reducing it |
| Routing or tender failure | A dense, pre-vetted carrier network so spot procurement is not the fallback | Rate negotiation on lanes where the problem is coverage, not price |
| Internal handling | Gate, dock and staging measurement — the material arrived, so find where it stopped | Attributing it to the supplier because they are the visible party |
| External | Alternative routings and carriers identified in advance of needing them | Treating every unexplained delay as external |
Two rows deserve a second look. Adding carriers to fix service failure frequently makes it worse — fragmented networks produce inconsistent service performance, because different carriers work to different on-time standards, exception protocols and damage thresholds, and for JIT and JIS operations that variability translates directly into expedite costs and line-down events that often dwarf the original transportation savings. And where the fix is coverage rather than price, negotiating the rate simply reduces the cost of a failure you keep having.
Was the Premium Justified?
A separate question from cause, and one almost nobody asks. Some expedites are correct decisions; others are over-servicing that erodes margin without protecting anything.
Justified
Coverage genuinely breached, with a production consequence that was avoided
The premium cost less than the disruption it prevented
Authorised by the named approver against a stated threshold
Recorded with the cause and the alternative that was rejected
Over-servicing
Expedited because it felt safer, with buffer still intact
Guaranteed window bought where a standard window would have held
Authorised informally, below any threshold that triggers review
Repeated on the same lane, which makes it a routing decision rather than an exception
Identify where premium services — expedited truck, express air, guaranteed windows — were used and ask whether those premiums were justified, because patterns of over-servicing erode margin without improving outcomes. A lane that expedites monthly is not experiencing exceptions; it is running the wrong standard service and paying a premium to conceal it.
Half your expedites are a planning number wearing a transport costume
Fleet Rabbit records each expedite event with its primary cause, named owner and justification, links it to the shipment and unit involved, and reports the split by category so the number reaches the function that can actually move it.
What to Report
Six figures. Note that total spend is deliberately not the headline — it moves with rates as well as with failures, which makes it a poor diagnostic on its own.
Expedite events per monthThe count, independent of cost. This is the planning signal, and it is the number that rate negotiation cannot flatter.
Premium as a share of freight spendAgainst the 3% to 10% benchmark band. Tells you where you sit and roughly what the recoverable gap is worth.
Split by primary causeEight categories, one per event. If any category exceeds half, check whether it is real or whether it has become the default selection.
Internal versus external attributionThe uncomfortable ratio, and the one that predicts whether the programme will change anything.
Justification rateShare of premium spend that met the stated threshold. Falling values indicate over-servicing rather than rising disruption.
Recurrence by cause and laneThe same cause on the same lane repeatedly is a design fault, not an exception — and it is the cheapest thing on this list to fix.
Frequently Asked Questions
What causes most premium freight?
Planning, not transport. Roughly 49% of expedite events are caused by inaccurate demand forecasts, which means nearly half of all emergency orders are the direct result of a planning problem rather than a logistics one. Add late plan amendments inside the frozen window and the internal share climbs further. That is why carrier negotiation rarely moves the total — it addresses the price of the symptom while the failure rate that generates the symptom stays constant.
How much premium freight is normal?
Top-performing supply chains spend around 3% of their freight budget on expediting; bottom performers spend around 10%. The gap between those two figures, multiplied by your annual logistics spend, is the recoverable opportunity — and it is recoverable rather than theoretical, because the difference is process discipline rather than market conditions. Establish where you sit before designing any programme, since the answer determines how much attention it deserves.
Why does attribution matter so much?
Because the department receiving the invoice is rarely the one that caused the cost, and the one that caused it never sees the number. Linking freight cost to the internal decisions that created it — late tenders, poor forecasting, order volatility, emergency shipments — reframes premium freight as a diagnostic of planning discipline rather than logistics performance. Presented that way it reaches people who can change the input; presented as a transport line it reaches people who can only negotiate the rate.
How deep should the analysis go?
Ask why at least four times. Stopping at "the supplier shipped late" attributes the cost accurately and identifies nothing actionable — the next question usually reveals that the call-off changed inside their lead time, and the one after that reveals an internal coordination failure. The corrective action that follows is a process change with a named owner and an escalation trigger, not an instruction to someone to try harder. That distinction is the difference between analysis and blame.
Will adding carriers reduce expedites?
Often the opposite. Fragmented networks produce inconsistent service performance because different carriers operate to different on-time standards, exception protocols and damage thresholds — and for just-in-time and just-in-sequence operations that variability translates directly into expedite costs, line-down events and premium charges that frequently dwarf the original transportation savings. What does help is a dense, pre-vetted network, since organisations without one face greater exposure to high-cost spot procurement when timing breaks down.
Should we ever question whether an expedite was needed?
Routinely. Identify where premium services — expedited truck, express air, guaranteed windows — were used and ask whether the premium was justified, because patterns of over-servicing erode margin without improving outcomes. The clearest tell is repetition: a lane that expedites every month is not experiencing exceptions, it is running the wrong standard service. Track justification rate alongside spend so over-servicing is distinguishable from genuinely rising disruption.
What is the first step?
Categorise three months of existing events retrospectively, using one primary cause each and no "supply chain" option. It takes a few days, needs no system change, and produces the split that determines everything after it. Most plants find the internal share is considerably larger than expected — which is the finding that gets planning and production into the room, and nothing else in this programme works until they are there. Start a free trial to record cause and owner against each event going forward.
Attribute It Where It Started
One primary cause per event, four whys before the answer is accepted, external requiring evidence, justification questioned as routinely as cause — and the number reported to the function that can actually change the input.
Works alongside existing TMS and ERP · Event-level attribution · Site-level configuration
August 20, 2026
By Emily Parker
All Articles
