Why measure maintenance
Maintenance teams are often asked whether things are getting better. Without numbers the answer is a feeling; with numbers it can become evidence. Key performance indicators, or KPIs, turn records of failures and repairs into a few figures that can be tracked over time. They work best as prompts for questions, not as grades. A number that moves invites the question “why?”, and the answer is where the learning is.
MTBF: mean time between failures
MTBF is the average operating time between one failure and the next for a repairable item. The calculation is straightforward:
MTBF = total operating time divided by the number of failures
Consider a hypothetical conveyor drive that ran for 4,000 hours in a period and failed 5 times. The MTBF is 4,000 / 5 = 800 hours. A higher figure suggests failures are less frequent.
Some care is needed with the words. Strictly, MTBF applies to repairable equipment. For items that are replaced and discarded, such as a light bulb, the related term is mean time to failure, or MTTF.
MTTR: mean time to repair
MTTR is the average time needed to restore equipment to service after a failure. Add up the repair durations and divide by the number of repairs. In the same hypothetical example, suppose the five repairs took 2, 3, 1, 6 and 3 hours. The total is 15 hours, so the MTTR is 15 / 5 = 3 hours.
What counts as repair time varies. Some teams start the clock when the machine stops, which includes waiting for a technician and waiting for parts. Others start when work begins. These give very different numbers, so the definition should be written down and used consistently. The same letters are also sometimes read as “mean time to respond” or “mean time to restore”, another reason to be explicit.
Availability
Availability is the share of time that equipment is in a condition to do its job. A simple version uses the two figures above:
Availability = MTBF / (MTBF + MTTR)
For the example, 800 / (800 + 3) = 800 / 803, which is about 0.996, or roughly 99.6 percent. This hypothetical figure ignores planned downtime and other real-world complications. The formula also shows why both numbers matter: availability can improve by lengthening time between failures or by shortening repairs.
Other common indicators
- Planned versus unplanned work. The share of maintenance hours spent on scheduled jobs compared with emergencies. A rising share of planned work is usually a sign of a calmer, more controlled operation.
- Backlog. The amount of identified work not yet done, often expressed in weeks of labor. A growing backlog may mean under-resourcing; a very small one may mean problems are not being recorded.
- Schedule compliance. The proportion of scheduled jobs completed on time.
- Preventive maintenance completion. Whether routine tasks are actually being done when due. Doing them is not the same as them being effective.
- Repeat failures. How often the same fault returns on the same equipment, which points to repairs that did not address the root cause.
- Maintenance cost. Often tracked relative to production output or to the value of the equipment, though comparisons between plants need caution.
Many of these tie into overall equipment uptime, a goal most plants share.
Limits and ways to be misled
Averages hide detail. In the example, the average repair time of 3 hours conceals one 6-hour repair, and that single event may matter far more to production than the others. Looking at the spread, not only the mean, is often more informative.
Other common problems:
- Small samples. Three failures in a year produce an MTBF that can swing widely. Treat numbers from few events as rough.
- Mixed equipment. Averaging the failures of very different machines blurs the picture. Calculate per asset or per class.
- Failures without records. MTBF depends on every failure being logged. If minor stops are fixed on the spot and never recorded, the figure looks better than reality. This is one reason work order quality matters.
- Constant failure assumption. MTBF treats failures as if they occur at a steady rate. Parts that wear out do not behave this way, and a single average can mislead about when failure becomes likely.
- Gaming. If a number is tied to rewards, people may close work orders early, avoid logging problems or define failure narrowly. Measures used for judgment tend to be distorted.
- Chasing a single metric. Pushing MTTR down by rushing repairs can increase repeat failures, which then hurts MTBF.
Targets taken from other companies or from published benchmarks are rarely directly comparable, because definitions and operating conditions differ. Trends within a plant, measured the same way over time, are usually more meaningful.
Choosing a small set
A reasonable approach is to choose a handful of indicators that match the plant’s current problems. A site fighting emergencies might watch the planned-work share and repeat failures. One with an expensive bottleneck machine might track availability and MTBF for that asset only. Review the measures regularly, and drop any that no one uses to make a decision.
Data comes most easily from maintenance software, and consistent coding of failures, causes and downtime makes the figures far more trustworthy. Where data is thin, even a simple log of date, equipment, stop time, restart time and cause can provide a start.
These figures are general and the example is hypothetical. Real calculations depend on how a plant defines operating time, downtime and failure.