Skip to content
allanninal.dev/build
Collaborate

Part 5 of 7 · Equipment maintenance scheduler series ~5 min read

What the history tells you

A year of maintenance records is the best available evidence about which equipment is worth keeping and which intervals are wrong, and it is almost always thrown away as closed work orders.

Key takeaways

  • Record what was found, including nothing. Nothing is the finding that changes intervals.
  • Cost and downtime by machine identifies the one worth replacing.
  • Deferral reasons in aggregate usually point at parts or people, not at scheduling.
  • Failures shortly after a service are worth counting separately.
  • Interval changes are deliberate, recorded decisions with the evidence attached.

Nothing is a finding

How three uneventful services lead to a proposed interval changeA vertical chain of five steps entered by a box labelled A completed service, asking what was found. Step one asks whether anything was replaced, parts or none; if none it exits to Record nothing found, explicitly. Step two counts how many times running nothing has been found. Step three asks whether it is three in a row for the same machine and job; if not it exits to Carry on, not enough evidence. Step four proposes a longer interval with the evidence attached. Step five has a person decide, and it is recorded. A note says this only works if nothing needed doing is a recordable outcome, which it rarely is.AWS ACCOUNTA completed servicewhat was found?Anything replaced?parts, or noneRecord 'nothing found'explicitlynoHow many times running?nothing foundThree in a row?same machine, same jobCarry onnot enough evidencenoPropose a longer intervalwith the evidenceA person decidesand it is recordedThis only works if 'nothing needed doing' is a recordable outcome, which it rarely is.
Fig 1. How intervals get lengthened on evidence. The first exit is the one most maintenance systems make impossible to record.
  • App integration
  • Machine learning
  • Management
  • Analytics
  • People

The recording problem

Most maintenance systems offer completed and not completed. A service where the engineer checked everything and replaced nothing is recorded identically to one where they replaced a worn belt, which throws away the single most useful signal about whether the interval is right.

Adding one field — what was replaced, with an explicit none option — makes interval optimisation possible. It is the cheapest improvement available to most maintenance operations.

The machine that keeps appearing

Planned and unplanned maintenance costs for four identical pressesA stacked bar chart with four bars in pounds per year. Two series: planned maintenance cost in green, and unplanned repairs and downtime in red. Press 1 costs one thousand two hundred planned and four hundred unplanned. Press 2 costs one thousand one hundred and fifty planned and three hundred and eighty unplanned. Press 3 costs one thousand three hundred planned and four thousand nine hundred unplanned. Press 4 costs one thousand one hundred and eighty planned and five hundred and twenty unplanned. A note says Press 3 costs more in breakdowns than the other three combined.£0£2000£4000£6000£8000~£1600Press 1~£1530Press 2~£6200Press 3~£1700Press 4Planned maintenance cost, £/yearUnplanned repairs and downtime, £/yearPress 3 costs more in breakdowns than the other three combined.
Fig 2. Four identical machines over a year. The planned cost is nearly the same for all of them; the unplanned cost identifies the one that should be replaced.

This chart is the business case for a replacement, and it is only possible because unplanned work was recorded against the machine rather than as a general repair cost. That linkage is worth insisting on.

The downtime cost is the harder half and is usually the larger one. Even a rough figure — hours stopped multiplied by a stated hourly value — makes the comparison possible, with the assumption written down so it can be argued about.

Failures after a service

Worth counting separately, because a cluster of them means the maintenance itself is introducing problems: a procedure that is wrong, a part that is not the right one, or reassembly that is not being checked.

It is an uncomfortable number to look at and it points at fixable things. A machine that fails within a week of a service twice in a year is telling you about the service, not about the machine.

What the report says

The quarterly page

  • Jobs due: 84. Completed on time: 61. Deferred: 19. Overdue, not deferred: 4.
  • Deferral debt: 37 machine-weeks past interval, up from 22.
  • Top deferral reason: part not available, 11 of 19.
  • Statutory: 6 due, 6 completed, 0 lapsed. This line must always read zero.
  • Unplanned cost by machine: Press 3 accounts for 61% of all unplanned work.
  • Interval changes proposed: 2, both lengthened, both with three clean services behind them.

The third line is the one that changes something outside maintenance. Eleven deferrals for missing parts is a stock-holding decision, and it is invisible until the deferral reasons are counted.

The fourth line is the one that should never be interesting. A statutory line that ever reads anything other than zero lapsed is the most important thing on the page, and putting it there every quarter is how it stays that way.

Next: what all of this costs to run.

All posts