Live data flowing in 7 days. No IT department required. See exactly what it looks like for your machines.
Calculate Your Downtime CostPractical guides, benchmarks, and shop floor strategies — no fluff, no spam, unsubscribe anytime.
Reactive vs preventive vs predictive maintenance for CNC shops. What signals to monitor, how vibration analysis works, real-world savings for a 5-machine shop, and common pitfalls to avoid.
Published OEE benchmarks are misleading because they mix industries. Here are CNC-specific benchmarks by machine type — turning, milling, multi-axis, Swiss — with realistic percentile ranges.
Lights-out machining adds 50-75% capacity at near-zero labor cost. But you cannot run unattended without monitoring. Here is what to monitor, how to build confidence in stages, and the ROI math.
If your OEE number is stuck below 60%, the answer is not “work harder.” It is almost certainly one or two of the Six Big Losses eating your productive time. The TPM framework gives you a name for every way a CNC machine wastes capacity. Once you can name it, you can measure it. Once you can measure it, you can fix it.
This article breaks down all six losses with specific examples from CNC machining. Not theory from a textbook — the actual things happening on your floor right now that are costing you money.
The Six Big Losses come from Total Productive Maintenance (TPM), the methodology developed by Seiichi Nakajima in the 1970s. They map directly to the three OEE factors: Availability, Performance, and Quality. Each factor has two loss categories, and every minute of lost production falls into exactly one of the six.
| OEE Factor | Loss Category | CNC Example | Typical Impact |
|---|---|---|---|
| Availability | 1. Equipment Failure | Spindle crash, servo fault | 5-10% |
| 2. Setup & Adjustments | Job changeover, fixture swap | 10-25% | |
| Performance | 3. Idling & Minor Stops | Chip clearing, material load | 5-15% |
| 4. Reduced Speed | Conservative feeds, worn tools | 5-10% | |
| Quality | 5. Process Defects | Scrap, rework, out-of-tolerance | 1-5% |
| 6. Startup Losses | First-article, warm-up cuts | 1-3% |
Impact percentages represent typical ranges for high-mix, low-volume CNC job shops. Dedicated production cells will skew differently.
This is the loss everyone notices because it is loud, expensive, and stops production cold. A broken tool holder. A servo drive fault. A spindle bearing that finally gives up after weeks of deteriorating vibration nobody was tracking.
In a CNC shop, equipment failures include:
Equipment failure typically accounts for 5-10% of total production time in a well-maintained shop. In shops running reactive maintenance only, it can hit 15-20%. The difference between those two numbers is the difference between preventive and predictive maintenance — catching the bearing at week 3 instead of week 11.
For most CNC job shops, this is the single biggest loss. Every time you change a job, the machine sits idle while the operator loads a new program, swaps fixtures, sets tool offsets, and runs a first article.
A typical job changeover on a 3-axis mill takes 30-90 minutes depending on complexity. If you run 3-4 changeovers per shift, that is 1.5 to 6 hours of an 8-hour shift spent not cutting metal. That is 19% to 75% of your planned production time — gone.
Setup loss is where SMED (Single-Minute Exchange of Dies) principles apply. The goal is not to eliminate setups — you run a job shop, setups are your business. The goal is to separate internal setup (machine must be stopped) from external setup (can be done while the machine runs the previous job). Pre-staging tooling, offline program verification, and standardized fixture systems all chip away at this number.
The first step is knowing how long your setups actually take. Operator-reported times are consistently 20-40% lower than machine- measured times. The machine does not lie.
These are the stops that nobody records because they are too short to feel significant. The operator clears chips from the vise. The bar feeder jams for 90 seconds. The coolant level drops and triggers a low-level alarm. The operator steps away to find a gage.
Individually, each stop is 30 seconds to 5 minutes. Collectively, they add up to 30-90 minutes per shift. And because nobody writes them down, they are invisible to management. You cannot fix what you cannot see.
Machine-level monitoring catches every one of these. When the spindle stops for 45 seconds, the system logs it. When it happens 14 times in a shift, you have a pattern. Maybe the chip conveyor needs adjustment. Maybe the bar feeder needs maintenance. Maybe the operator needs a better tool cart layout. The data tells you where to look.
This is the silent killer. The machine is running. Parts are coming out. But the machine is running at 70% of its programmed feed rate because the operator turned the override down three weeks ago and nobody noticed.
Common causes of reduced speed in CNC machining:
A machine running at 85% of its ideal cycle time all day looks normal from the floor. But across 5 machines and 250 working days, that 15% reduction is 1,500 lost productive hours per year. At $150/hr machine rate, that is $225,000 in capacity you already paid for but did not use.
Scrap and rework are the most expensive losses per unit because you have already invested machine time, material, and operator labor into a part that either goes in the bin or goes back to the machine for a second pass.
Most CNC shops run 95-99% quality on established jobs. But new jobs, difficult materials, and tight tolerances push that number down. And rework is often hidden — the operator catches a dimension at the machine, makes an offset adjustment, and re-runs the feature. That re-run time rarely gets recorded as a quality loss. It just disappears into the cycle time.
Tracking quality loss requires counting every part — good and bad — at the machine level. When you know that Job X produces 3% scrap on Machine A but 0.5% scrap on Machine B, you know where to look. Maybe Machine A needs a ballscrew replaced. Maybe the fixture is worn. The data narrows the investigation from hours to minutes.
Every time you start a new job or restart after a break, the first few parts are suspect. First-article inspection takes time. Warm-up cuts burn material. Offset adjustments require test passes. In aerospace and medical work, the first article alone can take 30-60 minutes of machine time.
Startup losses are often the smallest of the six in percentage terms (1-3%), but they compound with setup frequency. If you change jobs 4 times per shift and lose 15 minutes to startup each time, that is a full hour per shift — 250 hours per year per machine.
Reducing startup losses means standardizing your prove-out process. Same sequence, same gages, same first-article checklist. And machine data helps here too — if you can see the thermal profile of the machine at startup, you know exactly how long to warm up before cutting real parts.
Based on what we see across the shops we work with, here is the typical breakdown of where OEE points go missing in a high-mix CNC environment:
| Loss | Typical % of Total Loss | Easiest Win | Detection Method |
|---|---|---|---|
| Setup & Adjustments | 35-40% | External setup, pre-staging | Machine state monitoring |
| Idling & Minor Stops | 15-25% | Root cause analysis on patterns | Micro-stop logging |
| Equipment Failure | 10-20% | Predictive maintenance | Vibration + power monitoring |
| Reduced Speed | 10-15% | Feed rate override tracking | Spindle load + cycle time |
| Process Defects | 5-10% | SPC on critical dimensions | Part count + reject tracking |
| Startup Losses | 3-5% | Standardized prove-out | First-article time tracking |
The pattern is consistent: setup and changeover dominates in job shops, while equipment failure dominates in high-volume production. Your specific mix depends on your product variety, run lengths, and maintenance program. But you will not know your mix until you measure it.
See where your shop stands
Our free benchmark tool estimates your OEE breakdown by machine type and production style. Compare your shop to similar operations — no login required.
Run a free OEE benchmark for your shop →Do not try to fix all six at once. The playbook is simple:
This is not a 12-month consulting engagement. Shops that start measuring with machine-level data see their first actionable pattern within 2 weeks. The improvements start as soon as you can see where the time actually goes.
The Six Big Losses are not abstract categories. They are the specific reasons your $200,000 CNC machine is producing at 50% of its capacity instead of 75%. Equipment failures get the attention because they are dramatic. But in most job shops, it is setup time and micro-stops that quietly steal the most productive hours.
Name the losses. Measure them at the machine. Attack them in order of impact. That is the entire framework.
You cannot improve what you cannot categorize. The Six Big Losses give you the categories. Machine monitoring gives you the numbers.