1. Someone came to audit the shop today
This afternoon a robotics company came to audit us. They were here for about two hours.
What they build is not easy. Among the robot programs we have seen over the years, theirs sits in the hardest bracket — not “can it be made,” but “can it be made and then come out identical every time.”
One thing is worth saying up front: they did not come here to open the negotiation. Before the visit, they had already confirmed whether we could do the work. Only once the answer was yes did they arrange to come and look.
So these two hours were not about discovering anything. They were about verification. Look at the machines, look at the floor, watch how we do a first article, ask questions — four moves aimed at one thing: is what you said at the quotation stage actually true.
Later in the conversation, they told us plainly: this order is not an exception. Their parts list is always highly fragmented, and everything is always needed very fast.
2. “Many part numbers, needed fast” translates into one filter
What sat on the table at the quotation stage looked like this: 60 different parts, samples in 10 days.
What we only learned today is that this is simply their normal — not a one-off.
Most shops in our line hear that number and do two things. First they calculate a price. Then they decline.
What that filter removes is not the shop with the worst machines. Machines are purchasable — anyone willing to spend can have them. What it removes is the shop that only runs familiar work, only runs one part number in volume, and has never paid the price of a strange part.
Because the hard part of sixty SKUs is not how fast you cut. It is whether, before you quote, you can say which of those sixty will fight you, and at which operation.
If you cannot say it, you are working on luck. Quote enough jobs on luck and one of them lands.
3. The changeover math is not being re-run here
Why high-mix low-volume costs more is something we wrote about this year in Robot Joint and Flange Parts — changeover cost: new program, new fixture, new first article, the whole cycle every time the part number changes. We are not re-deriving that math today.
What we want to say is something else: being able to do the arithmetic is not the same as being able to take the job.
Knowing “sixty SKUs is expensive” and knowing “these sixty can leave here in ten days” are separated by craft, not by arithmetic.
4. We can talk about it because of the jobs that looked like a loss
A lot of what we take on does not look like good business at the time: one-offs, non-standard geometry, awkward tolerances, nasty materials, drawings missing half the information.
Slow, draining, and it pays less.
But there was one thing that never appeared on that ledger: every hard part we finished made our understanding of “hard” one degree more specific.
Specific enough that it stops being an adjective and becomes “how far this thin wall will spring in this material,” “whether this corner radius will chatter with this cutter,” “whether this tolerance survives three setups.”
None of that can be bought or crammed. It only comes from actually running the job, getting it wrong, fixing it, running it again.
So when that RFQ first landed on the table and asked whether we could do sixty parts in ten days, what let us keep talking was not nerve.
It was this: we have a fair idea of which of those sixty will hurt, and why.
With that, a quote is a quote. Without it, the price and the date are both bets.
As for whether we take the job, and whether we can actually deliver it — that is a separate question. The real answer is not in the meeting room. It is ten days out.
5. What an audit is really testing
What a customer can see on the floor is machines and housekeeping. Both are purchasable.
What they cannot see — but can draw out of you in half an hour of conversation — is something else: whether you will admit uncertainty when you are uncertain.
In this trade the dangerous answer is not “no.” It is “sure, we can do all of it.”
A supplier who agrees to everything tends to quote the lowest price and the shortest lead time, because he never worked out where the difficulty sits. When the goods come out wrong, the loss is not on him. It is on your line.
Why Quotes for the Same Part Differ covered one end of that: the invisible inspection cost that ends up transferred to the customer. This is the other end — whether the sentence was said clearly before the order was taken.
So these two hours were less an audit than an account reconciliation — taking what was said at the quotation stage and checking it, line by line, on the floor. Being willing to be checked is what makes it count.
6. Working with capable people is luck; doing it for years is selection
We took “long-termism” apart once already, in the Zhang Xue piece. That one was about being willing to book a visible loss in exchange for an invisible future.
Today, a second layer: what long-termism looks like when you are choosing who to work with.
It does not live in the contract terms. It lives where the contract does not reach.
For example: when a batch really goes wrong, is the other side’s first move to assign blame, or to fix it. For example: a small deviation that could be glossed over — do they raise it, or not.
Those two reflexes are hard to fake for long.
In a life, having one or two people who truly understand you is a rare thing. It is the same for a company.
I do not mean the kind you see every week and drink with. I mean the kind where you say half a sentence and they have the other half; where a decision gets made and nobody needs the reasoning explained again.
That kind of relationship cannot be bought with dinners. It grows one job at a time, each side declining to make the other look bad.
7. The trend is already visible; what is left is sharpening ourselves
Today’s audit is not an isolated event.
Many part numbers, needed fast, and still has to be right — that combination will show up more and more. The reason is simple: hardware people are moving from “one product, 100,000 units” to “ten products, 1,000 each,” and then to “dozens of products, a few dozen each.” The design is not frozen, the market is not validated, and nobody dares to bet on a big tool up front.
We wrote that side of it from the industry’s angle in The Components Boom. Today’s piece only adds the shop-floor view: what the trend looks like when it lands as one order.
For a machine shop, whether this change is good or bad comes down to one thing: do you treat variety as a nuisance, or as a craft.
Treat it as a nuisance and you keep picking the easy jobs, and your range narrows.
Treat it as a craft and you go looking for the hard ones, and your range widens.
We chose the second. Not for any noble reason — the logic in this trade is plain: how many kinds of hard parts you have made decides how hard a part people will trust you with.
For a few years I used to wonder, now and then, what a small shop like ours was really getting out of grinding through jobs that did not pay well.
Today I stopped wondering.
A team that builds some of the hardest robots in this country asked one question first — sixty different parts, ten days, can you do it — and only after hearing the answer did they decide to come and see for themselves.
A sentence came up in my head at that moment. Not aimed at the customer — aimed at ourselves:
Eternal CNC, are you ready?
We have been getting ready the whole time.
One line for you
Capability does not settle its account on the spot. It pays out all at once, on the day someone asks.
About Eternal CNC
We have made precision structural parts for twenty-three years. Among our thirty CNC machines we run a genuine five-axis, and we often take robot joint-module housings, reducer planet carriers and load-bearing structures — the kind where “a little off and it fails.”
We do not claim we can make anything. What we cannot make, we will tell you before the quote, not after.
If you have a batch of loose parts — many varieties, short deadline, awkward geometry — send the drawings. We will pick out the ugly ones first and tell you about them.