Total Productive Maintenance (TPM): what it is

Total Productive Maintenance —TPM— is a way of organizing maintenance that was born in Japanese industry and starts from an idea that’s uncomfortable for the classic model: the condition of equipment is not the exclusive responsibility of the maintenance department, but of the whole organization, from management down to the line operator.

The goal isn’t to repair better. It’s to eliminate the losses tied to downtime, quality and cost in production processes, and to do it continuously.

What counts as a loss

This is the concrete part of TPM, and the one worth understanding first, because it defines what needs to be measured. Production-system efficiency losses aren’t just breakdowns:

• Breakdowns, the ones everyone counts.
• Setup and adjustment, the time the machine is stopped while it’s being adjusted for another product.
• Tool or fixture changes.
• Cycle time deviations, when the machine is running but slower than it should.
• Minor stoppages, the two-minute ones nobody logs, which added up over a shift are the biggest loss of all.
• Reduced yield at startup, the ramp-up that takes time to stabilize.
• Quality problems: rework, touch-ups, scrapped product.

The list is useful because it explains how a plant can have few breakdowns and still have low efficiency. What isn’t measured doesn’t show up, and minor stoppages are almost never measured.

The pillars

TPM is usually described with several pillars. Four of them rest on the management system.

Autonomous maintenance. Production operators take on basic care of their own equipment: cleaning, visual inspection, lubrication, minor adjustments, parameter checks and anomaly detection. It’s the best-known pillar and has its own article, because it’s also the one that goes worst when it’s framed as a handover of work.

Planned maintenance. Classic preventive maintenance, with its routines and frequencies, carried out by the maintenance department. In a management system this means the preventive work orders that are generated automatically based on the frequency tied to the asset, its model or its family.

Focused improvement. Small groups tackling a specific, measured loss. The condition for it to work is having the data: without a track record, “improving” becomes opinion.

Quality maintenance. Linking equipment condition to product defects. It requires recording values, not just incidents.

What needs to be recorded

TPM is a methodology, not software, but there’s a part of it that doesn’t hold up without systematic recording. These are the four things you need to be capturing for the rest to make sense.

Stoppages, all of them. Including short ones. The work order in GMAO Cloud stores the cause of the stoppage, plus target, estimated and deviation times, which is what lets you compare what was planned against what actually happened.

Inspections, with values. A checklist of checkboxes tells you whether something was looked at; one with values tells you what was seen. In GMAO Cloud checklist fields support minimum and maximum values, so a reading out of range is logged as an anomaly on the spot, not as a side comment.

Anomalies detected by whoever operates the equipment. And, above all, that they reach someone: a logged anomaly that never becomes an incident with its priority and owner teaches the operator that logging things is pointless, and from then on they stop doing it.

Cost, per piece of equipment. Not the department’s cost, which is always documented. The cost of each asset, with its hours, its materials and its accumulated downtime.

The indicators

TPM is evaluated with data. The ones that come out of the management system itself and tell you whether something is changing are these.

MTBF, mean time between failures: measures reliability. If it goes up, the equipment fails less.

MTTR, mean time to repair: measures response capacity. If it goes down, we fix things faster.

Accumulated downtime per piece of equipment, which translates the above into lost production.

Ratio of preventive to corrective hours. This is the indicator that best describes whether an operation is proactive or reactive, and the slowest one to move.

Accumulated cost per asset, which lets you decide whether a machine gets repaired again or replaced.

The reports for MTBF and MTTR, downtime, anomalies, cost per equipment and hours per technician exist for this. But they’re only worth anything if someone recorded the times as they happened: a figure reconstructed at month’s end is no use for analyzing anything.

How it coexists with the maintenance you already do

A reasonable question is whether TPM replaces your existing maintenance plan. It doesn’t: it reorganizes it into layers, and each layer covers what the other one misses.

The daily layer is run by whoever operates the equipment, and its value is in frequency: it catches degradation before it becomes a breakdown. The planned layer is run by maintenance, and its value is in depth: replacements, measurements and checks that require expertise and resources. The corrective layer still exists, because breakdowns don’t disappear; what changes is that they stop being the main activity.

The sign that the layers are well distributed is easy to watch for: if corrective maintenance doesn’t drop after a few months on equipment with a daily route, either the route isn’t being followed, or it’s checking things that never fail.

Why it fails when it fails

Three reasons, and none of them technical.

It’s framed as a handover of tasks. If production believes maintenance’s work is being dumped on them, the initiative is dead before it starts.

The loop never closes. The operator logs something, and nothing happens. The anomaly never becomes a job, the improvement suggestion never gets a reply. Logging gets abandoned, and rightly so.

It isn’t measured. Without indicators, TPM turns into a set of good intentions and a cleaning round. The only way to defend it to management is by showing that downtime is going down.

Where to start

With one line or a group of critical equipment, not the whole plant. With its check routines defined, its frequency set, its autonomous-maintenance routes running, and its indicators measured from day one, so there’s something to compare against. Once that line has six months of track record, extending it is easy to justify; before that, it’s an act of faith.

If you want to see how the routines, frequencies and indicators are set up, you can request a demo.