Your container arrives. The parts are right, the welds are clean, and the finish matches the sample. You sign for it, mark the project “done,” and move on to the next SKU.
Nobody sends a thank-you note for a disaster that didn’t happen.
I want to walk you through one recent project to show you what you didn’t see. A buyer found a small metal fabrication factory on Alibaba. The factory looked sharp online, the communication was professional, and the reviews were solid. We got involved before the PO was placed and walked the factory floor with the buyer. It was a small team in a big workshop, with plenty of space for a bulky product. Everything checked out.
Then production started.
What a factory visit can and can’t tell you
A factory visit answers one question: do they have the equipment?
It does not answer the question that actually decides your outcome: do they have a system?
Equipment is something you can see. A system only shows up when something goes wrong. Here is what showed up over the next few weeks.
Surprise #1: There was nobody to read the drawing
Normally, we send the factory an English drawing. The factory’s engineer reviews it, converts it into a Chinese version the workers can read at a glance, and sends it back for us and the buyer to confirm.
This factory had no engineer, so there was nothing to convert.
So we drew it ourselves: a production-ready Chinese version, checked by us and confirmed by the buyer. If we hadn’t had in-house engineering, the options were to outsource the drawing to a third party or hand workers a drawing they couldn’t use.
Surprise #2: Nobody noticed the drawing contradicted itself
Drawings are not perfect. Sometimes two dimensions conflict, or one detail is ambiguous. A factory with real engineering capability catches this before cutting steel, with a simple message: “Page 2 says X, page 4 says Y. Which one is it?”
This factory caught nothing. The contradictions only surfaced after parts were made, which is the most expensive moment to find out.
Surprise #3: Everything depended on one worker’s line marking
There were no jigs or fixtures, and nothing to physically stop a worker from welding in the wrong place. Welding and assembly depended entirely on an experienced worker scribing lines and following them.
On the line, we found two welds in the wrong place. The worker’s marking was correct. Left and right were simply swapped. Every unit was supposed to be welded on the right side, and two had been welded on the left.
This is the problem with having no error-proofing: it puts the entire result on one person’s attention, on one afternoon. Everyone makes mistakes eventually, and good systems are built assuming they will.
Surprise #4: “We were just trying to help”
This is the one that should worry every buyer.
During the inspection, our QC found that the factory had changed the product without telling anyone. There was an opening in the design, and they had closed it off with a cover.
When we asked about it, the answer came without hesitation: “We saw an open gap, so we covered it for you. We were being helpful.”
They were sincere, and that’s exactly what makes it dangerous.
An unapproved change made with good intentions is still an unapproved change. That opening could have been there for ventilation, drainage, access, or compliance. It could have been there because the buyer’s customer specified it. The factory couldn’t know, and it didn’t ask.
If we hadn’t been on site that day, the unit would have gone into the carton, then the container, then the buyer’s warehouse. The buyer would have learned about it from a customer complaint or an Amazon review.
This isn’t about bad factories
I want to be clear about what this story is not.
This factory wasn’t dishonest. The people there worked hard, and they were not trying to cheat anyone. Most small factories are like this.
What they lack is a system around the skill. Many small factories are excellent at making things. What they don’t have is the layer between “customer sends a drawing” and “worker starts cutting”: someone to review, question, convert, and confirm. Without that layer, experience and intuition fill the gap. On a standard product, that works. On a custom product built to someone else’s drawing, intuition quietly becomes a quality risk.
Someone has to provide that layer. Either the factory has it, or your agent does.
The part nobody sees
Think about the last time you were healthy. You didn’t notice it or appreciate it. It was just how things were. You only understand its value after you get sick.
Sourcing works the same way.
When the product arrives on time and to spec, the buyer sees a normal outcome. They don’t see the drawing we rebuilt, the contradiction we caught, the two welds we flagged, or the cover we removed. The better the agent does the job, the less visible the job becomes.
That creates a strange trap: the agents who prevent the most problems are the easiest to undervalue.
So when you evaluate an agent, don’t only ask “what did you deliver?” Ask “what did you stop?”
What to look for
- In-house engineers. Someone who can produce and check drawings, and who is not a freelancer you’ll never meet.
- In-house QC inspecting during production. Final inspection finds problems after the work is done. Inline inspection finds them while they can still be fixed.
- Physical presence. An office and warehouse nearby, so someone can be on the floor the same day.
- A track record. Years of continuous operation mean they’ve seen these failures before and know where to look.
And one question worth asking any agent you’re considering: “What was the last serious problem you caught at a factory before it shipped?”
A good agent will answer immediately and in detail. If they can’t, they probably haven’t caught one, or they weren’t there to see it.
The best compliment a sourcing agent can get is silence.
About the author: Thomas is the founder of TOM Sourcing, a U.S.-registered sourcing and quality control agency with its own office and warehouse in China, serving brands in the US and EU since 2020.
