The $18,400 Mistake I Made Sourcing a Karmod Alternative for a US Project

The email came in at 4:42 p.m. on November 14, 2019. A facilities director at a Gulf Coast school district needed 12 relocatable classrooms on site by January 6 — the Monday before spring semester started. Not "around then." Not "ideally by." On site, inspected, and ready for class on day one.

Fifty-three days. That was the number on my whiteboard. I had already burned eleven of them going back and forth on a spec sheet I had written badly, and I still hadn't locked in a supplier. If you've ever managed a deadline where the client has zero flexibility, you know this feeling: the clock doesn't care about your process.

By the time the last module was finally placed in early February, we'd spent about $18,400 on port storage, rework, and one budget meeting I'd rather not relive. That's the number that taught me to stop comparing quotes the way I'd been comparing them for the previous two years.

How the decision started

We were sourcing modular classroom units — the typical 14 × 66 foot relocatable type, structural frame, full MEP rough-in. The district had used Karmod years earlier on a smaller project, so when we ran the RFQ, Karmod was the obvious benchmark. I sent the same spec package to six factories: Karmod, two other Turkish manufacturers, and three "alternative products" factories in Vietnam and China that my sourcing team was pushing hard.

Karmod's quote came back with clearly broken-out line items — production, FOB, sea freight, port handling, inland delivery. The Vietnam factory came back roughly 12% lower on the unit price alone. My team liked that number. My director liked that number. I liked that number, honestly, because on the face of it, it looked like a clean win.

That's when I made the first mistake. I compared unit prices and assumed the delta in every other category was roughly the same.

It wasn't. Turkey-to-Gulf-Coast is roughly three weeks of ocean transit in a normal year. Vietnam-to-Gulf-Coast is closer to four and a half, sometimes five, depending on routing. I knew this. I just didn't weigh it correctly against the 12% savings.

Mistake number one: "confirm lead time" without breaking it down

When I asked the Vietnam factory to confirm lead time, I got a one-line reply: "Production 6 weeks from PO." I took it at face value. I built a timeline around six weeks of production plus a four-week buffer, presented it to the district, and moved forward.

Here's what I didn't ask:

• Does the six weeks start from PO date, or from material in-factory? (Theirs: material in-factory. Steel delivery to the plant was another two weeks.)
• Who books the ocean freight? (My team, not theirs — a detail I learned by paying for it twice.)
• What does "production complete" mean — ready to ship, or loaded?
• What's the realistic port-to-port time in Q1, when Chinese New Year backs up everything for three weeks?

If I remember correctly, the first time we actually tracked the full sequence internally, the number came out to 12 weeks from PO to job site. The client-facing commitment I'd already made was nine. That two-week buffer I'd mentally allotted for "stuff happens"? It was gone before the first container hit the water.

Looking back, I should have insisted on a stage-by-stage schedule — in writing, with dates — and compared it against Karmod's equivalent schedule side by side. When I eventually did that comparison, the Karmod option's landed cost came within about $300 per unit of the "cheaper" alternative. The 12% savings wasn't 12% anymore. It was closer to 2%, and I was paying for the difference in days.

Mistake number two: writing "cold-formed steel" and thinking that was a spec

Our RFQ said "Cold-Formed Steel Residential Frame, per applicable US standards." That's a phrase, not a specification. It's the procurement equivalent of saying "a car, made of metal."

Two problems hid inside that sentence.

First, the thickness. CFS members are rolled from steel sheet measured in millimeters or in gauge. There's a real difference between 0.8 mm and 1.0 mm at the structural level — it affects the section properties in ways a structural engineer cares about more than a purchasing agent does. Our spec package listed "nominal 1.0 mm" on one page and didn't mention thickness at all on another. The factory read the second page.

Second, the coating. Per ASTM A653, galvanized sheet is specified by coating class — G60, G90, G115, and so on — measuring total zinc per square foot. The design requirements sit in AISI S100-16, the North American Specification for the Design of Cold-Formed Steel Structural Members. Our RFQ mentioned "galvanized" and nothing else. The factory defaulted to G60. The local building department wanted G90 for coastal exposure. Both are legitimate specs for their respective applications. One is wrong for this project. I'd assumed the two would align.

The real failure is that I didn't catch it until the framing drawings came back for review — three weeks into production, with containers already on the water. Our structural reviewer flagged the coating issue within about ten minutes of opening the file. The factory agreed to rework the affected panels, but the containers had already been loaded. That rework happened stateside, on our dime, at roughly $9,800.

The result, in numbers

The February closeout looked like this:

• Port storage and demurrage from the re-inspection hold: about $6,400
• State-side rework on the coating issue: about $9,800
• Inland transport re-sequencing when the rework slipped the schedule: about $2,200
• Total direct cost overrun: roughly $18,400
• Total schedule slip: 25 days past the January 6 deadline

The district was patient. The facilities director was patient. But we burned a relationship I'd spent two years building, and I still have that email thread saved in a folder called Lessons — Read Before RFQ.

What I changed afterward

A few process fixes stuck, and I'll share them because I genuinely think they'd help anyone working through a similar decision on a US project.

First, every lead-time confirmation gets broken into a stage table, not a single line. PO date → material at factory → production start → production complete → loaded at port → vessel departure → arrival at US port → clearance → inland delivery. Each stage gets its own number, and each number needs a source. "Confirm lead time" became "confirm the stage table."

Second, we don't compare unit prices anymore. We build a landed-cost sheet that includes freight, duties, port fees, inland delivery, and a time-value penalty for schedule risk. If a supplier can't or won't fill in the pieces, that's the signal.

Third, spec language got specific. "Cold-formed steel" became "CFS per AISI S100-16, minimum 1.0 mm nominal thickness, G90 per ASTM A653, with structural calcs stamped by a US-licensed engineer." The RFQ grew by a page and cut our rework by a lot more than a page's worth.

Fourth — and this is the one I wish someone had told me in 2018 — when a benchmark like Karmod is already in the picture, treat it as the comparison standard, not the thing to beat. Build the same landed-cost table for the benchmark, then hold every alternative to that exact table. If the alternative wins, great. But it should win on total landed, not on line item three of page two.

I'd like to say I got it right the first time. I didn't. But the second time, six months later, we ran the same comparison and caught the gap before the PO — $2,900 of additional freight cost that the unit price had completely hidden. The client never saw the save, and that's fine. That's the whole point of having a checklist. In our CivicAxis Sourcing workflow, the second version of that checklist now lives in a shared doc pinned to the top of the RFQ folder, and I update it every quarter whether anyone asks me to or not.