US rail intermodal volumes surge on truck-to-rail shift

The Analysis

A rail ramp is a batch machine bolted onto a road network that runs on a driver's clock. The ramp wants full blocks at fixed windows. The DC wants a live unload at eight in the morning. Everything that goes wrong in a truck-to-rail conversion happens in the gap between those two sentences.

The weekly rail data through week 25 puts US intermodal at 6,984,982 units year to date, up 3.3%. BNSF's second-quarter intermodal rose 9.5%, CSX 8.3%. The stated cause is price: high truckload rates and fuel pushed shippers to tender boxes to the steel. The same release adds, almost as an aside, that the volume is pressuring network velocity.

I want to walk one box through that conversion and find the step where the physical capacity runs out first. It is not the track. Track has slack. It is almost always the twenty miles at each end, and the paperwork that decides who pays for them.

Take the numbers apart before taking the shift apart. The 3.3% is a count of intermodal units, which means containers and trailers, moved year to date. It is not tonnage, it is not revenue, and it does not tell you how far each box went. BNSF at 9.5% and CSX at 8.3% are second-quarter figures, a different window from the year-to-date number, and the gap between 9.5 and 3.3 is itself information: the acceleration is recent, which means the operating problem is recent too. A network that has been carrying this volume for two years has learned where to put it. A network that picked it up in eight weeks has not.

Here is the physical sequence, and it has five steps. An import container is discharged at a coastal port, cleared, and drayed to a ramp. At the ramp it is grounded, then lifted onto a well car. The train runs to an inland hub. The box is grounded again, a chassis goes under it, a driver hooks up, and it goes to the DC. Two of those steps scale easily: a train can add cars, a terminal can add a track. Three of them do not. The lift at the origin ramp is a crane with a fixed cycle and a fixed number of operators per shift. The chassis pool is a finite stack. The dray move is a driver with an appointment, and that appointment sheet is the tightest capacity in the whole chain, because it is the only one measured in minutes.

Now the size of the wave. Divide 6,984,982 by 1.033 and last year's comparable figure is roughly 6.76 million, so the shift has added something like 223,000 units in 25 weeks. Spread that evenly over 175 days and it is about 1,275 extra boxes a day. That is my arithmetic on the published number, and the even spread is my assumption, not the railroad's. Nationally, 1,275 boxes a day is 53 an hour, which is noise. On one ramp it is not noise. If your ramp handles 5% of the national book, you inherited roughly 64 more lifts a day, 64 more chassis, 64 more gate transactions and 64 more dray moves. And a dray move is not a lift. It is a driver, a tractor, a slot on a sheet and a receiver who has to be open.

Who absorbs it depends less on your size than on whether you own the ramp interface. A big importer with dedicated blocks, its own chassis pool and a DC that runs two shifts gets through this with a phone call. A mid-size importer on a shared ramp through a 3PL gets the per diem, because the 3PL's appointment is not the one the terminal protects. A small shipper gets told there is no appointment until Thursday, and Thursday is the day the line needed the parts. Trade direction changes the answer too. Sell FCA at a US ramp and the conversion is somebody else's problem, your invoice does not move. Sell DDP with a delivered date on it and you have taken a linehaul saving and a service exposure on the ends, sitting in different budget lines, which is why nobody adds them up.

What the release cannot say is that a unit is a count, not a workload. A box going 300 miles and a box going 2,000 miles are both one unit in that 3.3%, so the number is loudest exactly where the network has slack, on the long haul, and silent about the ends, which is where the constraint lives. Underneath that sits the thing that actually moves your P&L. A modal shift does not remove variance. On the road the trucker absorbs the variance and prices it into the rate, so you see it. At a terminal the variance is stored: the box sits, the chassis clock runs, the appointment lapses, and the cost arrives later as per diem, detention and a missed receiving window. The total can still be lower. The risk is higher, and it is denominated in days, which is why it never gets compared with the dollars.

Timing, in the order you will feel it. Ramp appointment availability moves first, usually two to three weeks after a lane converts, and it is the only early warning that costs nothing to check. Chassis per diem follows at two to four weeks. Dwell shows up in the terminal's own numbers at three to five weeks. Your customer's on-time performance at the DC breaks at four to eight weeks, and by then you are arguing about a service failure rather than fixing a booking. The linehaul rate is the last thing to move and the first thing everyone watches.

Put my assumptions on the table, because they are mine and not the article's. Suppose you move 120 FEU a month into a Midwest DC and you converted 40% of it, so 48 boxes. Suppose ramp dwell rises by two days on each of them and chassis per diem runs about 45 dollars a day, an assumption and not a published figure. That is 48 times two times 45, or 4,320 dollars a month, before a single missed window. Add one expedited truckload a month at roughly 1,500 dollars to rescue a delivery date and you are near 5,800. If the rail linehaul saves 300 dollars a box, that is 14,400 dollars. Rail still wins. Roughly 40% of the saving has been eaten, and the part that ate it sits in a different cost centre from the part that earned it, so the person who approved the conversion never sees the bill.

The most common mistake in a partial conversion is treating it as a rate decision. It is an interface decision. Half your volume on rail means you keep a dray fleet relationship and a ramp relationship, and neither gets enough volume to earn priority at either. The second mistake is measuring the switch on the linehaul only, which guarantees you will expand a lane that is quietly losing money on accessorials. The third is assuming the ramp will tell you when it is full. It will not. The appointment sheet tells you, and only if somebody reads it every morning.

Run the other branch for a moment. If truckload rates and fuel had not pushed this volume across, ramp dwell stays where it was, chassis per diem stays nominal, and the drayage market stays soft enough that a missed appointment is a phone call instead of a day. The shift is not the problem. The problem is that it arrived faster than the ends of the chain could be resized, and the ends are the part with people in them.

What I would do this week is unglamorous. Get appointment wait times in hours for your three busiest ramps over the last sixty days. Get the per diem and detention line off your last three invoices. Put both next to the linehaul saving. If the accessorials are above a fifth of the saving, that lane is not converted, it is subsidised, and you either fix the interface or put volume back on the road until you can.

↑ Back to the key points

— By Ömer Kaya