← Insights

Maintenance

Preventive maintenance beats firefighting

July 2026 · 6 min read

Every maintenance department lives somewhere on a line between two modes. At one end, planned: tasks done on schedule, parts ordered in advance, downtime chosen to suit production. At the other, firefighting: the machine chooses the moment, the part is on express freight, and production stands watching a fitter work. The work is often the same work - the same bearing, the same belt. What differs is who picked the timing, and the cost multiplier that follows. A breakdown is just maintenance you did not plan, at triple the price plus the downtime.

Moving along that line does not start with software or sensors. It starts with a schedule that can express what your equipment actually needs, and a log honest enough to show whether you are winning.

Why month-based schedules quietly fail

Here is a small design decision with outsized consequences. Most home-built PM schedules compute the next due date by adding months, because the spreadsheet function for it is the one everyone knows. But maintenance does not happen in months. The lubrication round is weekly. The filter check is fortnightly. The guard interlock test might be every 30 days, the gearbox oil every 90, the full strip-down annual. A schedule that can only count in months cannot represent a weekly task at all - so the weekly tasks live on a separate paper sheet, the paper sheet stops being filled in around February, and the bearing that the weekly round would have heard grinding fails in April.

Intervals belong in days. 7, 14, 30, 90, 365 - one column, every real-world frequency, and next due is simply last done plus the interval. The spreadsheet convenience of months is not worth the tasks it cannot see.

With day arithmetic in place, the rest of the schedule mechanics follow: a live days-left countdown per task, amber at fourteen days so parts and access can be arranged, red at overdue. The colours are the to-do list; nobody should have to read dates to know what is due.

The two numbers that tell you if you are winning

The PM ratio - planned work as a share of all maintenance work - is the single best health indicator. Log every job with a type (preventive, corrective, inspection, breakdown) and the ratio computes itself. Below 50% you are firefighting; the schedule exists but the fires set the agenda. World-class plants run 80%+ planned. The number matters less than its direction: a PM ratio climbing quarter on quarter means the system is working, because every breakdown prevented converts triple-cost chaos into single-cost routine.

Downtime by asset - hours lost, ranked - is where the improvement budget should look. Downtime is never evenly spread; two or three assets usually own most of it. Ranked honestly, the list settles arguments that opinion never will: the machine everyone complains about is often not the one eating the hours. The worst actor might need a deeper PM regime, an operator-care routine, or a business case for replacement - but you only get to have that conversation if the log captures downtime hours on every job, breakdowns included.

Operators are the early-warning system

The cheapest maintenance capacity in the building is the person standing at the machine all shift. TPM formalises this as autonomous maintenance, but the practical version is simple: give operators a short daily-care routine - clean, check levels, listen, report anything odd - and make the checks visible. They catch the leak while it is a weep and the vibration while it is a hum. Those checks fold naturally into a layered process audit ("are today's TPM checks signed off?") and into the equipment pillar of a 5S programme - three systems reinforcing each other at the same workbench.

For ISO 9001 companies, all of this lands under clause 7.1.3: infrastructure must be determined, provided and maintained, with evidence. A schedule with computed due dates and a log with every job recorded is that evidence, produced as a by-product of simply running the system - which is the way compliance should always work.