Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- By 10 July, replace the network speed figure in your weekly report with two numbers from your provider: terminal dwell at your origin and destination ramps, and the share of trains held out of terminal.
- Re-quote every door-to-door commitment that includes a rail leg with a five-day spread instead of two, or book the difference as safety stock with a named cost per month.
- For the next two quarters, keep 10% to 15% of inland volume on the road as a variance valve, reviewed monthly against on-time performance.
- If your ramp is CSX-served, plan it as a separate lane with its own lead time and do not average it with a BNSF corridor.
- Write the trigger now: two consecutive weeks of rising dwell at your origin ramp moves the next two bookings to road without waiting for a meeting.
The US intermodal surge is dragging on network velocity: BNSF and Union Pacific intermodal train speeds fell to 10-month lows, Norfolk Southern is within 2% of a 20-month low, and CSX hit a seven-year low. Carriers including NS and CSX are hiring crew at dozens of locations to ease congestion.
The Analysis
There is one number in rail reporting that everybody quotes and almost nobody uses correctly: average train speed. It counts the hours a train spends standing still. That is the point of the metric, and it is also why a headline about a ten-month low deserves more attention than it got.
Through early July, BNSF and Union Pacific intermodal train speeds had fallen to ten-month lows. Norfolk Southern sat within 2% of a twenty-month low. CSX hit a seven-year low. Norfolk Southern and CSX are hiring crews at dozens of locations to ease the congestion.
A crew hire is a statement about how long the carrier thinks this lasts. Hiring is also the slowest lever in the yard. Between those two facts sits the two quarters your delivery promises have to survive.
The four carriers are not in the same trouble, and treating them as one market is the first mistake. A ten-month low and a seven-year low are different events. BNSF and Union Pacific at ten months means the deterioration is recent, which points at volume. CSX at seven years means something in that network was already tight before this freight arrived. If your inland leg runs over a CSX-served ramp, you are not in the same market as a shipper on a BNSF corridor, and a national average will mislead you in both directions: too calm if you are on the worst ramp, too alarming if you are on the best.
The mechanism is dull and entirely physical. More boxes on the steel means more train starts. More train starts means more trains arriving at a yard that was sized for last year's share. When the yard tracks fill, trains are held outside the terminal, and a train standing on the main line for six hours contributes six hours of zero miles to every average it touches. Speed is miles divided by total hours, and total hours include standing. That is why the number falls without anything dramatic happening: nothing broke, the queue just got longer at the top of the hill.
A small percentage here hides a large hour count there. If average speed drops 10%, transit time for the same distance rises by about 11%, because one divided by 0.9 is 1.11. That part is ordinary arithmetic and easy to live with. The part that is not easy is the spread. The mean moves a little; the tail moves a lot. A ten-month low in the average can easily be a two-year low at the ninetieth percentile, and appointments are made at the ninetieth percentile, not at the mean. Your customer does not receive the average container.
Layering matters more than usual. An importer running dedicated blocks or its own train gets a phone call and a rebooking. An importer sharing a ramp with everybody else gets whatever is left after the protected traffic moves, which is the wrong half of the distribution. Inland point intermodal cargo, the kind that rides a train to a Midwestern hub and then takes a dray, carries two terminal touches and therefore two chances to sit; port-only cargo that never leaves the marine terminal is barely in this story at all. Export direction flips the exposure. If you are loading at an inland ramp, the empty box has to arrive before your cargo can leave, and a slow network is slow in both directions, so the empty you were promised on Tuesday shows up on Friday and your cutoff was Wednesday.
On timing, the crew hiring is the most useful sentence in the release, and not for the reason it looks. Carriers do not hire at dozens of locations for a two-week problem. Hiring, training and qualifying takes months, not weeks, and that is my judgement about how railroading works rather than a figure from the article. So the carrier is telling you it expects this to run for a couple of quarters. Plan your service promises on that horizon. Improvements that show up sooner than that are noise or a local fix, not a trend you can bank.
The thing the release cannot tell you is that crew is the lever the carriers can pull at scale, not the lever that fixes this bottleneck. Put more crews on a railroad whose yards are full and you launch more trains into the same track; the queue moves off the crew board and into the yard, and dwell gets worse before it gets better. That is not incompetence, it is the only large lever available. It does mean you should not read a crew announcement as a service recovery date. The other thing it cannot tell you is that average train speed is cheap to compute and expensive to interpret. Ask for two figures instead: terminal dwell by yard, and the share of trains held out of terminal. Those two tell you where your box will actually be. The network average tells you what a spreadsheet did last week.
Here is the arithmetic with my assumptions written down, because none of these numbers come from the article. Say you move 60 FEU a month inland to a Midwest DC, which is about two boxes a day. Say your door-to-door promise is 22 days and the transit spread widens from plus or minus two days to plus or minus five. To hold the same promise you now carry three extra days of demand, which is six FEU. At an assumed 60,000 dollars of cargo value per FEU that is 360,000 dollars of extra inventory, and at an assumed 8% carrying cost it is about 28,800 dollars a year, before the warehouse charges you for the space. The alternative is to keep the inventory and break the promise, and one missed store delivery usually costs more than the inventory did.
The traps are predictable. Quoting a customer a date that the railroad's own dwell numbers cannot support is the big one, and it is always discovered by the customer first. Measuring yourself on the mean instead of the spread is the second. The third is over-correcting: pulling an entire inland lane back to the road restores control and gives back the linehaul saving you converted for, and it also means buying road capacity in a market that is already tight because everyone else had the same idea.
If the volume had not come across from the highway, none of this would be happening: dwell stays normal, speeds stay where they were, and a missed appointment is an annoyance rather than a planning input. The freight did come across, and the network will price that in one way or another. Your job is to decide whether you pay for it in inventory, in expedites, or in a broken promise, because standing still and hoping the speed number recovers is a decision to pay in all three.
↑ Back to the key points