CNC machine monitoring software reads data straight from each machine's control and shows you, as it happens, whether the machine is running, idle or stopped on an alarm, and how many parts it made. For a small job shop, its real value is showing where the hours actually go. On our two Fanuc VMCs, that was not where the owner thought.
We build and run FactoryOS, our own monitoring software. As of October 2026 it runs in daily production on two BFW VMCs with Fanuc controls at a CNC job shop. Everything below comes from that floor: what the owner learned, what setup takes, and what went wrong.
What does CNC machine monitoring software actually show?
At its simplest, three states and a count: run, idle, alarm, and parts made. A good system adds the detail you need to act on those numbers. In FactoryOS, that means:
- A machine-state timeline you can replay for any 24 hours
- Run, idle and alarm time, with utilization and availability per machine, and floor utilization
- Part and cycle counts, checked against the Fanuc part counter
- The running program, spindle speed and load, feed, and axis loads, live
- Average and best machining time per program, and the turnaround time from one part to the next
- Reports for today, yesterday, the last 7 days or a custom range
The difference from a register is who writes the number. A register or an operator's shift entry says what someone remembers or wants to report. The control has no reason to write off a long gap as "tool change".
| Production register | Operator entry in an app | Monitoring from the control | |
|---|---|---|---|
| Who records it | Supervisor, end of shift | Operator, when they remember | The machine, every cycle |
| Short idle gaps between parts | Usually missing | Often missing | Recorded |
| Part count | Written by hand | Typed in | Read from the control, cross-checked |
| Alarm history | Rarely | Sometimes | Every alarm, with time |
| Effort per shift | High | Medium | Low: check the gateway, keep load/unload times current |
How does it connect to a Fanuc control?
Fanuc controls talk to outside software through FOCAS. FANUC describes its FOCAS2 library as the way to communicate with a FANUC CNC over Ethernet or HSSB, with access to axis and spindle data, tool life, servo and spindle data, diagnostics, alarms and operator messages.
FOCAS can both read and write. In FactoryOS, a small gateway at each machine reads the control over Ethernet, read-only, and streams the data to our web app. Nothing is ever sent back to the machine.
Not every Fanuc control is ready out of the box. Before you talk to any vendor, walk to each machine and check these, based on MachineMetrics' FOCAS connectivity documentation:
- Control model. It lists Ethernet as standard on 0i-D/F, 16i-B/18i-B and 30i/31i/32i/35i controls. Older ones such as 0i-C, 21i-B and 16i/18i/21i-A can be upgraded by FANUC to include FOCAS, which is an extra cost and an extra wait.
- Ethernet port. CD38A is the most common port for FOCAS.
- IP address. The control needs a static IP, subnet mask and gateway on the same subnet as the edge device.
If your shop runs Siemens, Mitsubishi or Haas controls, FactoryOS is not the right fit today. It reads Fanuc only.
Where were the hours really going?
This was the part that surprised the shop owner. Before the data, the belief was that loading and unloading was the reason output was lower than it should be.
The timeline said something else. Most of the time, after a component finished machining and the cycle completed, the operator was not at the machine. The machine sat idle, ready for the next part, with nobody there to load it. That idle time could easily have been used to finish more components.
The owner then added up the idle time across all the months of data. It came to 35-40% more room for production on the same machines, with no new equipment.
What made that number trustworthy was a default loading and unloading time for each program. Every part needs some handling time between cycles. We set an expected load/unload time per program, so only the time beyond that counts as waste. Without it, every normal part change shows up as idle time and the waste figure is inflated.
As a separate feature, FactoryOS shows Hidden Time as a rupee cost at each machine's own hourly rate.
The lesson for any shop: don't trust an idle-time number until you know what allowance it gives for normal handling.
Should a small job shop chase an exact OEE number?
A full OEE figure needs more than machine states. It also needs to know how many parts were rejected and what the ideal cycle time for each part is. A small job shop usually doesn't have reject data tied to each machine, and ideal cycle times for one-off and short-run jobs are guesswork.
So FactoryOS shows an OEE-style view: utilization and availability per machine, and floor utilization, built from run, idle and alarm time it actually measured. We don't put a precise OEE percentage on screen, because it would be built on inputs we don't have.
Our advice: start with availability and idle cost. Those numbers are measured, not estimated, and they point at the biggest loss first. Add quality data later if your shop starts recording rejects per machine.
What does it take to start?
When a job shop owner asks us this, the answer is shorter than they expect:
- A small LAN connected to the machines.
- A gateway on the same network that collects the machine data and syncs it to our cloud. In our setup, that's a Raspberry Pi.
- The web app, which we host and run, with the backend and front end. Owners and supervisors open it in a phone browser; there is no separate app to install.
- Someone to set the basics: each machine's hourly rate and each program's expected load/unload time.
If the internet goes down, the gateway stores the data locally and syncs it when the connection comes back. You lose the live view for a while, not the record.
One thing that can slow a setup is in the checklist above: per MachineMetrics' FOCAS documentation, older controls need a FANUC upgrade to include FOCAS before any software can read them.
What went wrong for us
Three things caught us out; check for them in yours:
- The gateway sometimes stopped working or dropped off the network, and we didn't find out. Offline buffering protects you during an internet outage, but not when the gateway itself is down. Agree up front who checks it and how often.
- The internet line is a real discussion. Some shop owners question whether it's worth paying for a connection only for monitoring. Settle this before installation day, not after.
- Server cost is ongoing. Data goes to the cloud every day, so there is a running cost on top of the setup. Any vendor should tell you how that is billed.
We don't publish a fixed price or timeline for FactoryOS. Both depend on your controls, wiring and network, and we'd rather check those than guess.
If you want to see Industry 4.0 setups before you commit, the Ministry of Heavy Industries' SAMARTH Udyog Bharat 4.0 programme runs demonstration centres. The Center for Industry 4.0 (C4i4) Lab in Pune is the nearest for Maharashtra shops; the Smart Manufacturing Demo & Development Cell at CMTI is in Bengaluru.
How to choose CNC machine monitoring software
Skip the feature lists. Ask every vendor these questions:
| Question | Why it matters |
|---|---|
| Which controller models do you read today, with a live customer reference? | "Supports Fanuc" can mean one model. |
| Is the connection read-only? | You don't want software writing to your control. |
| What happens when the internet or the gateway goes down? | Buffering covers internet outages. Ask how they find out the gateway itself has stopped. |
| How do you separate normal load/unload time from real idle time? | Without it, the idle figure is inflated. |
| Is the OEE number measured or estimated? | Reject and ideal-cycle data have to come from somewhere. |
| Can I see it on my phone? | Owners aren't at a desk. |
| Who owns the data, and can I export it? | You may change vendors or connect an ERP later. |
| What is the ongoing cost, including server and internet? | The monthly bill matters as much as setup. |
When it's not worth it
If you have two or three machines and you or a trusted supervisor stand next to them all shift, you already see most of what monitoring would show. If your controls are old and need an upgrade on every machine, the upgrade cost may come before any saving.
Monitoring earns its keep when machines run while nobody is watching every cycle. In our two-machine case, that's where the 35-40% came from.
Machine data is one piece. Most shops also need job-work, stock and dispatch in one place. See how we approach software for manufacturing units, and what to look for in inventory management software for manufacturers. If you need something built around your own floor, our custom web application and ERP development starts from the same questions.