ODM CNC Machining Parts: A 7-Step Fanuc CNC Pre-Production Checklist (Simulator Included)

I run pre-production checks at a contract machine shop. The work is simple to describe, although not always simple to execute: customers send drawings, and we return machined parts. In the trade, it's called CNC machining parts ODM—parts made to the customer's design, shipped under their label. I've handled this kind of work for eight years. I've also personally made and documented six significant mistakes, totaling roughly $41,000 in wasted material, rework, and rushed replacements.

After the third rejection in Q1 2024, I started maintaining a written pre-production checklist. It has seven steps, and it's written around Fanuc CNC controls—not because Fanuc is the only brand that works, but because it's what we run on most of our machines.

When To Use This Checklist

We run it before these situations:

• The first article of a new part number.
• A repeat order after more than a few months.
• Any program edited by hand or generated from a new CAM post-processor.
• A setup that will be run by an operator who hasn't seen the fixture before.

If it's a simple job you've made 500 times, skip what feels unnecessary. But for anything with a customer deadline attached, the full seven steps take less time than explaining a rejected batch.

The 7-Step Checklist

Step 1: Confirm the actual Fanuc controller CNC model and version

“It’s a Fanuc” isn’t enough information. A Fanuc controller CNC series can be a 0i-MF on a vertical mill, a 0i-TF on a lathe, or a 30i/31i on a multi-axis machine. Those controls don’t behave identically. Machine builders also enable different options, so even two mills running the same controller model can use different M-codes and features.

Press SYSTEM on the control and read the series and software version. Write it on the setup sheet. That 20 seconds tells you which post-processor profile and simulator setting you should trust in the later steps.

Step 2: Match the CAM post-processor to the machine, not just the controller

A “generic Fanuc” post-processor can produce legal-looking G-code—until it doesn’t. The controller executes commands, but the machine builder decides how those commands interact with the tool changer, pallet changer, rotary table, spindle orientation, and coolant system.

Here’s what I mean: a horizontal machining center and a vertical machining center can share the same Fanuc controller model, but their kinematics are different. A post built for one can produce a program that drives the other machine in the wrong direction. The CAM backplot won’t necessarily show that, because the backplot uses the CAM model, not the machine’s geometry.

I only believed this after using the wrong post on a 4-axis job and watching the first article come out as a mirror image. That mistake was a $2,100 rush remake and a three-day delay. Now we name post files by the exact machine asset number, not by the controller family.

Step 3: Compare the setup sheet against the offsets in the controller

This step doesn’t need a simulator. Print the setup sheet and put it next to the control. Check one thing at a time:

• Units—G20 inch mode or G21 metric mode.
• Work coordinate values—G54, G55, etc.
• Tool numbers and tool length offset numbers.
• Fixture offset values if the machine uses external work offsets.

I made the classic version of this mistake in my first year, 2017. I loaded the right tool but copied the wrong tool length offset from the previous job. It didn’t crash; it just cut a pocket about 0.5 mm deeper than the drawing allowed. By the time anyone measured it, 20 parts had run. That order went from profitable to a loss.

Now every program header on our floor starts with G21 or G20 written explicitly. Those commands are modal, meaning the control can still be in the mode left by the previous job if the new program doesn’t reset it. Make the program say what it expects.

Step 4: Run the exact G-code file through a Fanuc CNC simulator

Most people search for “fanuc cnc simulator free download” because they want to practice button presses without standing in front of a machine. That’s a fair reason. But the simulator didn’t become part of our production checklist for training. It became part because it catches a different layer of errors than CAM software.

The CAM backplot shows the toolpath the software thinks it created. It doesn’t always simulate what the control will actually do with modal G-codes, macro variables, subprogram calls, and machine-specific settings. Running the exact file through a Fanuc-style simulator executes those commands and raises alarms when something doesn’t exist or doesn’t make sense. CAM can say the part is fine while the control says otherwise. Why do we check both? Because they catch different mistakes.

Everything I’d read about CNC simulators said they were for classrooms, not production. My experience with Fanuc-controlled machines suggests otherwise. We once had a post that output an arc with start and end points the control rejected. The CAM simulation drew the arc perfectly because it recalculated the geometry from the CAD model. The simulator raised an alarm before the machine could move. That saved us a scrapped first article on an ODM order that had a two-week lead time.

One honest warning: there isn’t a single official “Fanuc CNC simulator free download” that works with every controller model. Ask your Fanuc distributor which simulator or demo version matches the controller from Step 1. Avoid cracked downloads from random forums. A simulator that doesn’t match your control version can give you false confidence, and false confidence is worse than no simulator.

Step 5: Do a dry run with the fixture on the table

The simulator doesn’t know what’s actually clamped on your table. It doesn’t know about the clamp that’s two millimeters above the programmed path, or the vise jaw that was left in the wrong position after the last job.

A dry run should be boring. We put the real fixture on the table, set the work offset, and run the program in AUTO with single block on and rapid override at a low setting. We watch the tool approach each feature for the first time. No coolant, no cutting, no rush. If something doesn’t look right, we stop and ask why.

In September 2022, I told a new setup tech to “just do a dry run” while I took a phone call. He set the rapid override to 100% and stepped away for a moment. The program rapid-traversed directly into a clamp that wasn’t in the CAM model. That mistake wrecked a $1,400 fixture and cost us a week of schedule. The simulator would not have caught it, because the simulator doesn’t know where we placed the clamps.

Step 6: Measure the first article before you run the rest

After the machine has proven it can move safely, let it make one part. Then stop. Don’t start the batch. Measure that first part against the drawing.

I know it’s tempting to keep the spindle running while the part is still in the vise. But the first article is not the time to trust a caliper reading on a tight tolerance. Anything at ±0.05 mm or tighter goes to a proper bore gauge, micrometer, or CMM before we release the rest of the order.

Here’s how I learned it. In Q1 2024, we had a bore that looked fine, felt fine, and seemed fine to everyone on the floor. It was 0.02 mm outside the upper tolerance. The customer’s incoming inspection caught it after we had already delivered the full 120-piece order. $2,800 straight to scrap, plus a one-week delay and a very unhappy buyer. That rejection was the third in four weeks, and it’s the reason this checklist exists.

If you’re doing ODM work, remember that the finished part carries your customer’s name. They don’t want to hear about setup issues. They just want parts that pass.

Step 7: Archive the as-run program with the order

Once the first article passes, copy the final program out of the control and save it with the order number, part number, and drawing revision. Don’t overwrite the old version. Keep the setup sheet, the offset values, and the tool list in the same folder.

This step matters more than most people think. Operators often make small edits during dry runs or first-article setup. If you don’t save the as-run version, the file on your server and the file in the machine will be different. A month later, when the next batch comes in, someone loads the old server file and repeats the same manual fix. Or worse, they don’t repeat the fix, and the program runs wrong.

We’ve caught 47 potential errors using this checklist in the past 18 months. Most were caught in Steps 1–4 before a spindle ever turned. Some were as simple as a mismatched tool offset. Those are the easiest errors to fix. They’re also the easiest to miss when you’re in a hurry.

Reminders We Add To Keep This Checklist Honest

First, the most expensive phrase in our shop used to be “it looked fine in CAM.” CAM simulation and control simulation are not the same thing. One verifies the ideal toolpath. The other verifies what the machine will actually execute. Use both.

Second, this checklist isn’t about Fanuc being superior to other controls. It’s about knowing your exact machine. The same logic applies if you run Siemens, Heidenhain, Mitsubishi, or anything else. Know the model, know the post, know the offsets, prove the program, and measure what comes out.

Third, the same vagueness applies to laser work. Some buyers say “just laser cut it” and expect the same edge quality no matter which machine runs it. In our shop, a fiber laser handles most of the steel sheet work; a CO2 laser is still the right tool for certain non-metals and edge-critical jobs. Neither is automatically better—they’re different. If your ODM order includes machined parts and laser-cut parts together, put the real requirement on the drawing: material, thickness, and edge condition. “Laser cut” alone leaves too much open.

One Scope Note for CO2 Laser Searches

If you arrived here wondering “how long after co2 laser can you wear makeup,” please follow the instructions from the doctor who performed your procedure, not from a machinist. The CO2 lasers we work with are industrial cutting tools. The laser used in skin resurfacing shares the acronym, but it is a completely different machine and a completely different recovery process.

This checklist is meant for people who produce industrial parts. Print it if it helps. Keep it next to the controller. And take it from someone who paid $41,000 for the reminder: the 30 minutes of checking is almost always cheaper than the rejection that follows the shortcut.