This article is for plant managers, maintenance managers and continuous improvement leads who have launched, or are about to launch, an autonomous maintenance program and need to know, with numbers, whether it is working.
- Judge an autonomous maintenance program on results, not on checklists completed.
- Five weekly indicators are enough: route completion, abnormalities found, short stops per shift, unplanned downtime, availability.
- Short stops are the earliest result signal, but they rarely appear on paper logs.
- Expect findings and backlog to rise first: that is a sign the inspections work.
Autonomous maintenance looks simple on paper: operators take on cleaning, inspection, lubrication and small adjustments, and maintenance technicians get time back for planned work. In practice, most programs are judged on feelings. The routes are walked, the checklists are signed, and six months later nobody can say whether the line runs better than before. This is where measurement earns its place.
What autonomous maintenance is supposed to change
Autonomous maintenance (AM), sometimes called operator-based reliability, is one of the pillars of Total Productive Maintenance. Its promise is not “operators do maintenance”. Its promise is earlier detection: a loose guard, a leaking fitting or a worn belt gets noticed by the person who sees the machine every hour, before it turns into a breakdown or a string of small stops.
So the outcome to measure is not the number of checklists completed. It is whether abnormal conditions are caught earlier, and whether the machine loses less time as a result. Those are two different signals, and they do not move at the same time.
Lessons from IDCON on running an AM program
IDCON, the reliability and maintenance consultancy based in Raleigh, North Carolina, has published one of the most practical guides on the subject: How to Implement Autonomous Maintenance, by Torbjörn Idhammar. It walks through the whole work system: defining operator tasks, documenting inspection routes in walking order, keyword checklists, measuring points, the backlog that grows when inspections start finding problems, and the role of leadership.
Two pieces of advice from that guide matter directly for measurement. The first is about restraint: “Do not collect more measuring points than the plant can handle and have time to analyze.” The second is a warning: “A good way to kill an AM/OBR process is to ignore failure reports and collected data.”
Both point to the same thing. Data only helps an AM program if it is small enough to be read and if someone visibly acts on it. Operators who report a problem and never see a response stop reporting. The guide also explains that backlog and maintenance costs tend to rise before reliability improves, which is exactly why leadership needs a way to see early progress.
Five indicators that show whether AM is working
A practical AM scorecard does not need dozens of KPIs. Five indicators, read every week, are enough to tell the difference between a program that is taking root and one that is drifting.
| Indicator | What it tells you | Where the data comes from | What to expect early on |
|---|---|---|---|
| Route completion rate | Whether operators actually walk the inspection routes | Checklists, CMMS | Should stabilize within a few weeks |
| Abnormalities found per route | Whether inspections detect real conditions | Inspection reports, work requests | Rises first: inspections find what was always there |
| Short stops per shift | Whether small, repeated losses are going down | Automatic machine stop data | Falls once the first findings are fixed |
| Unplanned downtime (hours per week) | Whether breakdowns are being prevented | Automatic stop data, CMMS | Falls more slowly than short stops |
| Availability and OEE on the pilot line | The overall effect on productive time | Automatic machine stop data | Moves last, once the other indicators improve |
The first two indicators measure the activity. The last three measure the result. A program that only tracks the first two can look healthy on paper while the machine behaves exactly as before.
Why short stops are the earliest result signal
Breakdowns are rare events. On a single line, a few prevented failures over a quarter are hard to tell apart from luck. Short stops are different: a jam, a misfeed or a sensor fault can happen dozens of times per shift, so a change in their frequency shows up within weeks.
They are also the losses most directly linked to what operators inspect and clean. A dirty photocell, a worn guide or a loose fixture rarely causes a breakdown, but it causes repeated two-minute stops that the operator fixes alone and never writes down. That is the problem: on paper logs, short stops are mostly invisible. They show up as a vague gap between the expected output and the real one.
This is where automatic stop data complements the IDCON approach. When every stop is timestamped from the machine itself, the AM team can see, line by line and shift by shift, whether the stops tied to the inspected components are going down after the abnormalities are fixed.
A 90-day measurement plan for an AM pilot
The following sequence keeps the data small, in line with IDCON’s advice, and makes progress visible to operators and leadership.
- Weeks 1 to 2: baseline before changing anything. Record every stop on the pilot line automatically, with its duration and time of day. Do not reclassify anything. This gives an honest starting point for short stops, unplanned downtime and availability.
- Weeks 3 to 4: launch routes and log findings. Start the inspection routes and record each abnormality found. Expect the number of findings and the maintenance backlog to rise. Share that rise as a good sign, not a failure.
- Weeks 5 to 8: close the loop. Fix the findings that match the most frequent short stops first. Show operators, at the line, which stops went down after their reports. This is the feedback that keeps the program alive.
- Weeks 9 to 12: compare against the baseline. Compare short stops per shift, unplanned downtime and availability with the first two weeks. Decide which routes to extend to the next line and which checks can be dropped.
The visual below shows the typical pattern on a pilot line: findings rise first, then short stops fall. The data is illustrative, not from a specific plant.
Common measurement mistakes
- Counting checklists as results. A 100 % route completion rate says nothing about the machine.
- Measuring only breakdowns. They are too rare to show progress in the first months, so the program looks like it is not working.
- Relying on hand-written stop logs. Short stops are the first to disappear from paper, which hides the very losses AM targets.
- Collecting data nobody reviews. As IDCON puts it, ignoring collected data is a good way to kill the process.
- Changing too many lines at once. Without a clear pilot and baseline, improvements cannot be attributed.
With TEEPTRAK, automatic stop capture on a pilot line takes about one hour of installation per machine: a module wired in the electrical cabinet and a sensor at the edge of the line, with no connection to the PLC. OEE and stop data are available within 48 hours, which is enough to build the baseline before routes start. Across plants using TEEPTRAK, unplanned stops fall by 30 % on average.
Further reading
- How to Implement Autonomous Maintenance, Torbjörn Idhammar, IDCON: the full work system, from route design to leadership commitment.
- Autonomous maintenance: giving operators the data to spot drift before a breakdown
- From Reactive to Predictive: TPM with Real-Time Data
FAQ
What are the best KPIs for autonomous maintenance? Track both activity and results: route completion and abnormalities found for the activity, then short stops per shift, unplanned downtime and availability for the results. Five indicators read weekly are usually enough.
How long before autonomous maintenance shows results? Abnormalities found usually rise in the first weeks. Short stops are typically the first result indicator to fall, once findings are fixed. Breakdowns and availability move more slowly, over several months.
Why do maintenance costs rise at the start of an AM program? Inspections uncover conditions that were already there, which temporarily increases the backlog. IDCON describes this rise as expected, and leadership should plan for it before reliability improves.
0 Comments