What TPM actually is
Total productive maintenance is not a maintenance strategy in the way that preventive or predictive maintenance are — those are ways of deciding when to intervene. TPM is a plant-wide operating model: it says that equipment effectiveness is everybody's job, that operators are the first line of defence rather than the people who report faults, and that the score is kept in one number — OEE.
That framing matters, because it explains why TPM programmes stall. A plant can buy condition monitoring, write a preventive schedule and build a dashboard without changing who feels responsible for a machine. TPM without that shift in ownership is just maintenance with more meetings.
The eight pillars (and the foundation)
The pillars are conventionally drawn as a temple standing on 5S — sort, set in order, shine, standardise, sustain. The foundation is not decoration: if a workplace has no standard for where things live and what clean looks like, an operator cannot see that something is wrong.
| Pillar | What it means in practice |
|---|---|
| Autonomous maintenance | Operators clean, inspect, lubricate and tighten their own machines to a standard — and are trained and trusted to. This is the pillar that changes outcomes. |
| Planned maintenance | Interventions scheduled on time, cycles or condition instead of on failure. See predictive versus preventive maintenance for how to choose. |
| Focused improvement (kaizen) | Small cross-functional teams attacking one named loss at a time, with before-and-after numbers. |
| Quality maintenance | Setting and holding the equipment conditions that produce good parts, so quality stops being an inspection problem. |
| Early equipment management | Feeding maintenance experience into the specification and commissioning of the next machine, so you stop buying the same problem twice. |
| Training & skills | Building the competence the other pillars assume — the pillar most often skipped, and the reason autonomous maintenance so often fails. |
| Safety, health & environment | Zero-incident intent built into every standard, not audited in afterwards. |
| TPM in administration | Applying the same loss-elimination thinking to planning, procurement and stores — where a great deal of plant downtime is actually created. |
The six big losses — TPM's actual target
TPM's objectives are usually stated as zero breakdowns, zero defects and zero accidents. The working version is more useful: eliminate the six big losses, which map directly onto the three components of OEE.
- Availability losses: breakdowns, and setup & adjustment (changeovers).
- Performance losses: small stops and idling, and reduced speed.
- Quality losses: start-up and yield rejects, and production rejects/rework.
This is why TPM and OEE are inseparable in practice. OEE is not a companion metric to TPM — it is the scoreboard TPM plays on. If you cannot see which of the six losses is largest this month, you are choosing improvement projects by opinion. Our OEE calculator will do the arithmetic on one shift's numbers, and MES versus OEE covers where each system belongs.
How to actually start: six steps
- Pick one line, not the plant. A pilot on the bottleneck gives you a result in months instead of a programme that dies in year two.
- Establish a baseline you believe. Measure OEE on that line for four to six weeks with reason-coded stops before changing anything. Without a credible baseline, every later claim is contestable.
- Do 5S properly on that line. Not a poster campaign — a defined place for tools and spares, a defined clean state, and a standard someone owns.
- Start autonomous maintenance small. One cleaning-and-inspection standard, one lubrication route, done by operators, checked weekly. Cleaning is inspection: it is when a leak, a loose guard or a hot bearing gets noticed.
- Attack the largest loss with a kaizen team. One loss, one team, a measured before and after. Publish the result whether it worked or not.
- Only then extend. Second line, then planned maintenance and quality maintenance pillars. Extending before the first line holds its gains just distributes a failure.
The South African complication: TPM under load shedding
A programme built on stable rhythms meets a grid that isn't. Three adjustments make TPM survivable here:
- Code power losses separately. If load shedding sits in the same bucket as breakdowns, your availability trend becomes noise and TPM teams end up chasing a loss they cannot fix. A distinct reason code keeps the controllable losses visible — the same argument we make in the cost of unplanned downtime.
- Use the outage window. A scheduled shedding slot is planned time that already exists. Plants that move lubrication routes and inspections into it recover hours that were being lost anyway.
- Count restart losses honestly. Repeated restarts add start-up rejects and thermal stress. Both are TPM losses, both are measurable, and both usually go uncounted.
Where TPM sits among lean manufacturing tools
TPM is one tool in the lean set, not a rival to it. 5S is its foundation; SMED (single-minute exchange of die) attacks the setup and adjustment loss; standard work stabilises the performance loss; root cause analysis is what a kaizen team does when it stops guessing; visual management is how the result stays visible. Value stream mapping sits above all of them and tells you which line to point them at. A plant does not need all of these to start — it needs one line, one measured baseline and one loss worth attacking.
The honest failure mode: most TPM programmes we hear about did not fail because the model was wrong. They failed because the plant installed dashboards without changing who owns the machine, or ran the pillars without ever measuring OEE — so after eighteen months nobody could prove anything had changed, and the programme quietly lost its budget. Measurement without ownership stalls; ownership without measurement cannot defend itself.