6:40 p.m., Wednesday, November 12, 2025
The call hit my cell, not the office line. That's never a good sign.
I coordinate rush orders at Rodcrest Sourcing, which is a polite way of saying my phone rings at night. The buyer on the other end ran a fleet rebuild shop in the Midwest — Cummins ISX and Cat C15 work, mostly. They had 24 diesel engine connecting rod assemblies sitting on a bench and a machine shop telling them the rods they'd ordered two weeks earlier were the wrong weight code. Blocks already torn down. Six trucks parked. Their contract carried a $2,500-per-day late clause, and the collection for the freight bound to the assembly floor was Friday at 4 p.m.
"I know it's short notice," he said. "I know."
Thirty-six hours.
What we had, and what I assumed
We had 24 forged connecting rods in stock for that platform. 4340 steel, standard weight, small-end bushings already pressed. That part was lucky.
What wasn't lucky: the stock came from two separate production runs, and the earlier one had been sitting on a rack nobody had touched since 2023.
Here's where I should have stopped and written something down. Instead I told the buyer, "Standard spec, same as what you ordered. We'll build the assemblies tonight and they'll ship by noon tomorrow."
Now the part I have to be honest about. The buyer had sent over a spec sheet. On it, under crankshaft journal size, were the words .010 under. But the purchase order line item read "standard bearings, matched set." His own shop had ground the crank and nobody had updated the PO.
I went with the PO. I didn't cross-check it against the spec sheet, because the order had come through clean and this buyer and I had done a dozen jobs together. That's not a defense. That's just what happened.
The 1:20 a.m. detail
Our assembly lead, Dana, caught it at 1:20 in the morning. She was setting up the bearing press and opened the box of connecting rod bearings we'd pulled — standard wall thickness — and asked the question I should have asked twelve hours earlier: "These are standard. Is the crank standard?"
It wasn't.
If you've ever done engine assembly, you know what happens when a standard-wall bearing goes into a rod feeding a .010-under journal. The oil clearance goes to nothing. You spin a bearing at best. At worst you're pulling the engine back out of a truck in six weeks and explaining it to somebody's service manager.
If you've done this work, you know that feeling in your stomach.
So we stopped. Twenty-four rods, four of them half-built with bearings already pressed in, and a decision to make at 1:30 in the morning.
The fix, which almost failed twice
Three options on the table.
One: ship the standard bearings and let the shop handle the clearance. Non-starter. That's just handing the problem back to a customer who was already behind — put another way, it's not a solution, it's a delay with paperwork.
Two: source .010-under bearings overnight and press them ourselves. That's what we did, but it burned four hours. Our usual bearing supplier doesn't stock that thickness for this platform, and we called three distributors before we found a set. The connecting rod bearing isn't a part most people think about until it's the only thing that matters.
Three: swap the whole order to a billet connecting rod assembly built from solid stock. We looked at it seriously. Billet connecting rods are a genuinely good answer for high-RPM or boosted builds where you control the entire spec envelope. For a fleet C15 running near stock power, forged 4340 was already the right call, and building from billet stock would have taken nine days. Not a real option.
Then the second problem surfaced. Dana measured the big-end bore on all 24 rods, because at that point nobody was trusting anything on that rack. Twenty of them sat within 0.0003 of each other. Four did not — they came from the 2023 run, and the bore ran about 0.0005 larger. With a thicker bearing wall you have less room for error, not more.
We honed those four to match. Cost: $940, and roughly two hours — actually closer to three when I look at the shop timestamps.
I knew I should have pulled a sample off both racks and measured before I confirmed the order. I thought, what are the odds? The odds turned out to be 4 in 24.
2:10 p.m., Friday
The assemblies left at 2:10 p.m. Friday, under the wire. The forwarder collected at 3:40. My total extra cost: $820 in expedited freight, $940 in rework, and one extremely uncomfortable phone call with the buyer.
We ate most of it. The client stayed.
What would have happened if we'd shipped on time with the wrong bearings? Best case, their shop catches it at assembly and eats another two days. Worst case, a truck goes down a month later and both of us are sitting in a conference room with somebody's warranty department.
What I actually changed
Three things. None of them are glamorous.
The spec sheet wins, always. If the PO line item and the drawing disagree, the drawing wins, and the order stops until somebody confirms in writing. I don't think that's a novel idea — every sourcing person reading this already knows it. Knowing it and doing it at 6:40 on a Wednesday evening when a buyer sounds desperate are two very different skills.
Samples from every rack. Two batches of "the same part" are not the same part. Same supplier, two years apart, and the tolerance bands have drifted while nobody noticed. You check the sample or you gamble.
We started logging the corrections. We'd never tracked how often rush orders turn out to need a spec fix. Since January 2026 we've logged it. Out of 63 rush orders this year, 19 needed a spec or part correction before shipping — about 30%. Nobody books a rush order expecting a paperwork problem, but the numbers say it's roughly a coin flip whether the documents and the hardware actually agree.
On the premium, since people ask
I have mixed feelings about rush fees. On one hand, a $3,150 expedite on a $14,000 order feels like a shakedown. On the other hand, I've now stood in a shop at 1 a.m. watching Dana re-measure a rack of rods because the paperwork lied — that premium isn't arbitrary. It's the price of flexibility the original schedule never accounted for.
At least, that's been my experience with deadline-critical work in this category. We've only run this kind of correction on three C15 batches so far, so I'd stop short of calling it a rule.
Take it from someone who has now run enough of these that the pattern is boring: in 2020 we'd have gotten a faxed spec sheet and a phone call. In 2026 we get PDFs, occasionally a 3D scan of a crankshaft journal, and the same mistakes still happen. The fundamentals didn't change. The execution did.
The best practice from 2020 doesn't fully apply now — but measuring the part in front of you was good practice then, and it's still good practice today.