Preventive Maintenance Scheduling in a Machine Shop: Treating PM as a Capacity Block
Preventive maintenance that lives on a separate calendar collides with the production schedule. This guide shows how to book PM as a capacity block that removes hours before jobs are loaded.

Why PM collides with the production schedule
It is 6:50 on a Tuesday morning and the supervisor of a two-shift CNC shop is scanning the board. The 4-axis horizontal on the second line has a rush housing loaded from 7:00 straight through to the end of second shift, and the job looks fine. Then the maintenance lead walks over with a clipboard. The monthly spindle and coolant service on that same machine is due today. It was written on the maintenance calendar three weeks ago, on a sheet the scheduler never sees.
Two choices, both bad. Pull the machine for four hours and the rush job slips past its ship date, or skip the service one more time and hope the spindle holds until the weekend. Nobody did anything careless. The production schedule promised hours that the maintenance calendar had already spent.
This is the usual failure of preventive maintenance scheduling in a machine shop: PM lives on one calendar, jobs live on another, and the machine is double-booked before anyone loads a single part. The fix is not a better reminder. It is to treat every PM window as a capacity block that removes hours from the machine before jobs are loaded. The rest of this article shows how, with a worked example you can rebuild for your own machines.
The collision has a simple cause. A job schedule answers the question "what runs when?" A maintenance calendar answers "what gets serviced when?" Both are claims on the same scarce thing, spindle hours, but only one of them is usually visible to the person promising delivery dates. When the claims live in separate places, the first one to be noticed wins, and it is almost always the job with the customer waiting. PM gets deferred, shortened or squeezed into a weekend, and the deferral is invisible until something fails.
Preventive maintenance scheduling in a machine shop: the capacity-block rule
The rule is one sentence: a PM window is not a task on a to-do list, it is hours the machine does not have.
If a machine is booked for service, those hours are gone from the schedule before a single job is loaded.
That reframes PM from an interruption into an input. In finite-capacity terms (see finite capacity scheduling explained), a machine can hold one job in one window, and a PM block simply occupies a window the same way a job does. The difference is when it is placed. Jobs are loaded into whatever hours remain; PM is subtracted first.
Three consequences follow from doing it this way.
- Available hours become the number you plan against. Planned shift hours minus PM blocks is the capacity the schedule may promise. Anything loaded beyond it is a commitment the machine cannot keep.
- Conflicts surface at planning time, not on the floor. If a rush order wants a window that a PM block already holds, you see the clash on the board while there is still room to move something, instead of at 6:50 with a clipboard in your face.
- PM stops being the first thing cut. A block that is visible on the same board as the jobs has to be argued with explicitly. A task on a separate sheet can be skipped silently.
None of this requires software to understand. It requires one place where both claims are written down, and a habit of subtracting before adding.
Worked example: one machine, one month
This is an illustration with round inputs, not a measurement from any shop. Take a single CNC machine on two 8-hour shifts, five days a week.
- Scheduled hours per week: 2 shifts × 8 hours × 5 days = 80 hours.
- Scheduled hours over a four-week month: 80 × 4 = 320 hours.
Now add the PM the machine needs, again using round numbers:
- A 2-hour weekly service (lubrication, chip clearing, coolant check) × 4 weeks = 8 hours.
- One 8-hour monthly block (spindle, way covers, filters) = 8 hours.
- PM total for the month: 8 + 8 = 16 hours.
Available hours to promise against: 320 − 16 = 304 hours.
If the scheduler loads jobs against 320 hours, the schedule overpromises by 16 hours, which is two full shifts. Those 16 hours do not vanish. They come back as one of three things: a late job, an overtime shift, or a PM task that quietly gets skipped. All three cost something, and the third is the one nobody can see until the machine says so.
Notice what the arithmetic does not say. It does not say PM is expensive. At these inputs it is 5% of scheduled hours (16 ÷ 320). It says that a plan built on 320 hours was never a plan the machine could keep. Your numbers will differ, so for your own preventive maintenance scheduling in a machine shop, substitute your own shift pattern and your builder's service intervals. The method is the point.
A related note on measurement: how planned maintenance time is treated inside OEE depends on the convention a shop adopts, so decide it once and apply it to every machine. The OEE calculation guide walks through the availability, performance and quality terms so you can see where each choice lands.
What unplanned downtime costs, and why it is the wrong comparison
The usual argument for PM is the cost of the failure it prevents. Two published figures are worth knowing, as long as they are labelled for what they are.
Siemens' "True Cost of Downtime 2024" report looked at the world's 500 largest companies and put the figure at roughly 25 unplanned incidents and 27 hours of downtime per month for a large plant, about 326 hours a year (Siemens/Senseye, 2024, via vsight.io). Those are large-plant numbers. A 20-machine job shop should not read them as a forecast for its own floor.
Separately, the ABB Value of Reliability survey of 3,215 plant maintenance leaders reported a median cost of about $125,000 per hour of unplanned downtime, across industries (ABB, 2023, via Acronis). That is a cross-industry median, and it should not be applied to a single machine shop as if it were your hourly rate. The two figures come from different samples and measure different things, so they are shown side by side here and never averaged.
What both really establish is direction: unplanned stoppages are frequent and expensive at scale. For your own shop, the better calculation is local. Take the hours a failure would have taken from a machine and price them against the margin on the jobs that machine would have run. The sibling guide on the cost of unplanned downtime in a job shop shows how to build that number from your own records.
But cost of failure is the wrong thing to lead with when you are scheduling. The scheduling question is smaller and easier to answer: how many hours does this machine have this week, once the maintenance it needs has been taken out?
Building a CNC preventive maintenance calendar the schedule can read
A PM calendar for CNC machines only helps the scheduler if it is expressed in the same units as the schedule: a machine, a start, an end. A reminder that says "service the lathe in March" is not a capacity block. A block that says "Lathe 3, Thursday 14:00–16:00" is.
A workable build has four steps.
1. List every machine and its real working hours. Name, type, cell, shift pattern and active status come first, because a block can only remove hours that exist. A machine on one shift has a different ceiling than one on three. If your machine list is incomplete or its shift data is stale, the available-hours arithmetic above will be wrong from the start. The machine capacity planning guide covers how to set that baseline.
2. Take intervals and durations from the builder, then adjust from history. Spindle, lubrication, coolant, filtration and way-cover tasks all come with intervals in the machine's manual. We do not publish a one-size interval table here because it varies by machine, duty cycle and material. Start from the manual, then tighten or relax the interval using your own failure and inspection records.
3. Convert each task into a dated block. Weekly tasks become short recurring blocks, monthly tasks become longer ones, and annual or calibration work gets booked well in advance. Place them where they hurt least: the end of a shift, a known low-load window, or a gap between jobs. A block that sits in a natural gap costs nothing; one that cuts across a running job costs a changeover and a restart.
4. Subtract before you load. Compute available hours as shift hours minus blocks, then load jobs against that number only. When a rush order arrives, it competes for what is left, and moving a PM block becomes a deliberate decision with a named owner rather than a drift.
Two cautions from the domain. First, keep blocks short and frequent where you can; many small windows are easier to absorb than one that takes a machine out for a day. Second, record when a block is moved or skipped. A schedule that shows deferred PM is telling you something about real capacity, and a pattern of deferrals on one machine is a signal to staff or reschedule differently, not to stop booking PM.
Keeping the blocks honest
Preventive maintenance scheduling in a machine shop only works if the capacity block survives contact with the floor. Three habits keep it honest.
Review the next two weeks of blocks in the same meeting that reviews the jobs. If PM is discussed in a separate meeting, the two calendars will drift apart again.
Close the loop. When a block finishes early, release the unused hours back to the schedule. When it runs long, extend the block and re-check what sits behind it. Either way, the schedule should reflect what actually happened.
Treat a skipped block as a capacity event. If a service is skipped to protect a job, log it and rebook it on the board. A deferred PM that is not rebooked is just deleted maintenance.
You can build all of this in a spreadsheet today. The Machine PM & Capacity-Block Scheduling Workbook (/store/machine-pm-capacity-block-scheduling-workbook) lets you schedule planned and preventive maintenance windows per machine, then reserve each one as a capacity block that reduces that machine's available hours. If you have not yet catalogued your floor, the Machine Registry & Capacity-Block Setup Workbook (/store/machine-registry-capacity-block-setup-workbook) catalogs every machine on your floor, with name, type, cell, shift pattern, hourly capacity and active status, and is the natural place to start.
Either way, the next step is small. Pick the one machine that gets rushed most often, write down its real shift hours, subtract the PM it needs this month, and see what number is left. If that number is lower than the hours your schedule currently promises, you have just found the collision before it found you.
Ready to go beyond the guide?
Most shops are on a live Gantt board within 60 minutes of sign-up, with their existing job list imported from Excel.
Get shop floor scheduling guides in your inbox
Practical articles for production managers — no spam, unsubscribe anytime.
Related articles
Scheduling Around the Bottleneck Machine: Drum-Buffer-Rope for Job Shops
An hour lost at the bottleneck is an hour lost for the whole shop. Here is how to identify the constraint machine and bu…
CNC Setup Time Reduction: A SMED Walkthrough for Job Shops
In a high-mix shop, setups eat the capacity the schedule thinks it has. This walkthrough applies SMED to CNC changeovers…
Machine Utilization Benchmarks by Machine Type: Mills, Lathes, and EDM
A shop-wide utilization number flattens out machines that behave nothing alike. Here's how mills, lathes, and EDM typica…