Downtime Pareto: find where your machine hours go
A downtime Pareto sorts stop reasons by total minutes lost, so the few causes behind most lost hours stand out. Collect every stop with start, end and a short reason code, sum minutes per reason over a week or month, sort from largest to smallest and attack the top two or three with specific actions, then check the next Pareto to see if they shrank.
Guilherme Rodrigues ItinoseUpdated 2 min read
Key takeaways
- Sort by minutes lost, not by number of occurrences.
- Keep reason codes short: eight to twelve options, plus 'other'.
- Micro-stops add up; capture them automatically if possible.
- Act on the top two or three causes, then measure again.
- Review the Pareto weekly with the people who can fix the causes.
Every plant has a list of things that “always stop the machine”. A Pareto replaces that list with numbers.
Build it in four steps
- Capture every stop with start, end, machine and reason.
- Sum minutes per reason for the period.
- Sort from largest to smallest and add the cumulative percentage.
- Pick the top causes that make up most of the lost time.
A useful set of reason codes
| Code | Example |
|---|---|
| Setup / changeover | New program and fixture |
| Tool change / breakage | Insert or drill replaced |
| Waiting material | Bar stock or blanks missing |
| Waiting inspection | First part waiting for QA |
| Maintenance / breakdown | Hydraulic leak, spindle alarm |
| No operator | Break, meeting, absence |
| Program or offset adjustment | Corrections during the run |
| Other | Free text, reviewed weekly |
Turn bars into actions
For each top cause, write one concrete action with an owner and a date: a setup kit by the machine, a material call in the ERP, a pre-set tool cart. Next week’s Pareto tells you if it worked.
CNC Pulse builds the loss Pareto automatically from FANUC controls. Our automation team can also connect reasons to tablets or WhatsApp.
When to outsource this
- You know machines stop but not why
- Improvement projects are chosen by opinion instead of data
When not to
- You have no stop data at all yet; collect it first
FAQ
Minutes or occurrences?
Minutes show where time goes; occurrences show annoyance. Use minutes to prioritize, occurrences to spot micro-stops.
How many reason codes?
Few enough to choose in seconds. Long lists end up with everything under 'other'.
What about 'unknown' stops?
If unknown is a top bar, fix data collection first: a quick reason prompt or rules based on alarms and programs.
How long a period?
A week shows quick issues; a month smooths noise. Keep both views.
Send us one drawing
We return a review with issues found, suggested fixes and a fixed-price scope.
