Why I Was the Wrong Person to Source an SMT Production Line

I want to be upfront about something before I tell this story: I'd never sourced a full SMT production line before. I'd bought component placement machines. I'd ordered a couple of reflow ovens. That's not the same thing. Not even close.

I'm a capital equipment procurement guy handling electronics assembly orders for about 9 years now. My experience is based on roughly 40 equipment purchases, most of them standalone machines. If you're working with a mature SMT setup already and this is just another upgrade, your situation might look nothing like mine.

OK, here's what happened.

In February 2022, we signed a contract that required us to bring SMT assembly in-house. Until then, we'd been outsourcing all surface-mount work to a partner shop. The client wanted us to own the process. My director gave me a $380,000 budget and 16 weeks.

I said yes on the spot. Which, in hindsight, was probably my first mistake.

The "Obvious" Plan (That Wasn't)

My background was in purchasing standalone machines. When you buy one piece of equipment, you do exactly one thing: you compare unit prices and specs. That's the muscle I had. So that's the muscle I used.

I built a spreadsheet. Loader, printer, two placement machines, reflow oven, AOI, unloader. Six categories, three vendor quotes each. I chose the lowest price per category that met our throughput spec. Total came in at $342,000. I felt great.

Here's what my spreadsheet didn't have:
- SMT production line layout (how the machines physically fit, and where feeder carts could be staged during changeover)
- SMT production line setup (installation, calibration, first-article runs, process recipe development)
- PCB assembly integration (how the SMT line talks to our existing through-hole and test stations)

I didn't know to include those things, because nobody had ever told me they mattered. The vendors quoting me weren't going to bring it up either — each one only quoted their own piece of the puzzle. That's the trap with SMT production line sourcing done as a shopping list: every quote is technically correct in isolation, and the whole thing is still broken.

"A procurement spreadsheet can compare machines. It can't compare whether those machines actually work together on your floor."

The Day the Second Delivery Arrived

The equipment landed at the end of June — three days late, but whatever. We had a crew scheduled. Everything got uncrated. And that's when the first problem showed up.

The placement machines had been sized for a 12-inch conveyor width. The printer's outfeed was 14 inches. Not a dealbreaker in itself, but it meant we needed an adapter section that nobody had quoted. And the adapter section, once we priced it, meant the reflow oven's inlet height was now off spec.

We didn't catch any of this until the machines were already bolted down.

What followed was a week of back-and-forth that I still cringe about. The printer vendor blamed the placement vendor. The placement vendor blamed the oven vendor. The oven vendor sent us a PDF. I sat in the middle of it, on the phone, at 11pm, trying to figure out who was technically responsible for a conveyor height.

Even after we got the mechanical stuff sorted, the line still didn't produce. First articles failed AOI at three times the acceptable rate. Reflow temperature profile was off. Our IPC-A-610 Class 2 target for the new line wasn't even close.

We brought in a process consultant in week 11. He walked the floor for about two hours, then pointed at three things:

  1. The placement machines were mounted too close to the reflow oven. In-line staging buffer was essentially zero.
  2. The feeder layout on the high-mix side was mirrored from the recommended orientation — meaning changeover time was double what it should've been.
  3. The AOI station was downstream of the unloader, which sounds trivial until you realize we were sampling from the wrong end of the process.

None of these were broken machines. Every machine was doing exactly what its spec sheet said it would do. The SMT production line layout just hadn't been designed as a system.

What It Actually Cost

Let me put numbers on it, because this is the part I want people to remember.

The adapter section and rework: $8,200. The consultant (six days, including travel): $11,400. The temperature profile development, which took nine days instead of the two we'd planned for: roughly $6,800 in lost production time. The later conveyor swap, when we finally admitted the layout was wrong: $14,600. And the client penalties for the two-week late start: $5,900.

Round that to $47,000 over budget, on a line we'd already bought "at the lowest price per category."

I approved every one of those invoices. Hit "confirm" on the first one and immediately thought: did I miss something in the quoting stage? I didn't relax until the first shipping run went out under the Class 2 IPC-A-610 standard, about five weeks later.

For what it's worth, the whole mess probably would have been caught in 30 minutes by anyone who had done exactly this job before. I hadn't. That's the whole lesson.

What I Changed Afterward

I now run SMT production line sourcing as a structured process, not a spreadsheet exercise. A few things I do differently:

  • Ask vendors what they don't do. The vendor who says "we do placement and reflow, we don't touch line layout — here's someone who does it well" is worth ten vendors who claim to cover the whole thing. Every time. That's not a marketing line. It's the difference between a working line and a $47K lesson.
  • Budget line layout and setup as separate line items. Not embedded in a quote, not "included." Separate. If it's not a line item, it isn't real.
  • Get the layout on the floor before the POs go out. Tape or paint the machine footprints on the actual floor. Walk it. Note where a person has to stand during changeover. It sounds stupid. It catches things.
  • Define the PCB assembly integration touchpoints up front. How does the SMT line hand off to through-hole? To test? To conformal coat? Those interfaces are where the months go.

(Note to self: also stop agreeing to "16 weeks" before I've met the vendors.)

The Bit That Took Me Longer to Accept

The real mistake wasn't on the spreadsheet. It was on my side of the table.

I'd assumed my job was to buy machines at the best price, and someone else's job was to make them work together. Nobody told me that. Nobody has that job assigned by default. On a system as interconnected as an SMT production line, the integration is the project. The machines are just the parts.

Now I write that assumption into the plan from day one. And I look for a partner who says it out loud. If a sourcing vendor tells me "we're great at equipment, we're not a layout specialist," I trust their equipment recommendations a lot more. Weirdly, that's the signal I use now.

That $47,000 came out of a budget I'd rather have spent elsewhere. But we've since run 23 more equipment sourcing projects, and I've personally caught about $180K worth of would-be mistakes using the checklist I built after that February. So, you know. Net positive. Eventually.