Troubleshooting

Maintenance and Downtime Metrics: Common Mistakes and How to Catch Them

The most common errors in reliability and downtime numbers, from mixing planned with unplanned downtime to counting operating hours wrong, each with a symptom and a fix.

The most frequent mistake is lumping planned downtime into unplanned failure metrics. Symptom: your MTBF looks terrible, say 40 hours, while the machine rarely breaks. Root cause: scheduled changeovers, PMs, and lunch stops got counted as failures in the MTBF tally. Fix: only count unplanned stops in the failure count, and route scheduled stops through the Planned Downtime Percentage tool separately. If you logged 12 unplanned failures over 480 operating hours, MTBF is 40 hours, not the 8 hours you get when 48 changeovers pollute the failure count.

Unit and time-base errors wreck availability figures. Symptom: Equipment Availability reads 98 percent but the floor feels like it stops constantly. Root cause: the denominator uses calendar hours (168 per week) instead of scheduled run time, hiding stoppages inside idle time you never planned to run. Fix: define the time base once and hold it. If you scheduled 120 hours and lost 18 to breakdowns, availability is (120 minus 18) divided by 120, which is 85 percent, not the inflated number a 168-hour denominator produces. Document whether you mean loading time or calendar time.

Double-counting or missing repair segments distorts MTTR. Symptom: MTTR of 22 minutes when technicians swear jobs take over an hour. Root cause: the clock starts at wrench time and ignores wait-for-parts and diagnosis. Fix: decide whether MTTR is pure repair or total restore time, then apply the MTTR tool consistently. A job with 15 minutes diagnosis, 30 minutes waiting on a bearing, and 20 minutes wrench time is 65 minutes if you measure restore, or 20 if you measure repair only. Mixing the two definitions across assets makes fleet MTTR meaningless.

Averaging availability across dissimilar assets produces a false plant number. Symptom: plant availability of 90 percent that no single line matches. Root cause: a simple mean of five machines ignores that one bottleneck runs 80 hours a week while others run 10. Fix: weight by scheduled hours or throughput. If your constraint runs at 82 percent and the rest at 95 percent, the number that matters for output is the constraint, not the arithmetic mean of 91 percent that flatters the plant and hides where the losses actually sit.

Stale or wrong labor and rate inputs silently corrupt downtime cost. Symptom: Downtime Cost per Hour looks reasonable at 1,200 dollars but finance disputes it. Root cause: the model uses last year's contribution margin and forgets idle labor still on the clock. Fix: rebuild the rate from current lost margin plus burdened labor. A line making 240 units per hour at 9 dollars margin loses 2,160 dollars per hour in margin alone, before adding 6 operators at 38 dollars burdened, pushing true cost past 2,380 dollars per hour. Refresh these inputs quarterly.

Ignoring short stops and micro-stoppages understates real losses. Symptom: your logged downtime is 4 hours a week but scrap and late orders say otherwise. Root cause: stops under a threshold, often 2 minutes, never get recorded, yet 30 such stops a shift add 60 minutes of hidden loss. Fix: capture minor stops via machine PLC counts, not manual logs. Fifteen unrecorded 90-second jams per shift equal 22.5 minutes, roughly 5 percent of an 8-hour shift, enough to move Equipment Availability from a claimed 96 percent to an actual 91 percent once counted.

Confusing Equipment Availability with Maintenance Availability sends improvement effort the wrong way. Symptom: you invest in faster repairs but availability barely moves. Root cause: your losses are dominated by waiting and admin delay, not repair speed, so the maintenance-availability view was the right lens. Fix: split the two. If MTBF is 50 hours and MTTR is 5 hours, inherent availability is 50 divided by 55, or 91 percent, but if logistics delay adds 10 hours of wait per event, operational availability drops to 50 divided by 65, near 77 percent. Attack the delay, not the wrench time.

Using too small a sample makes reliability numbers swing wildly. Symptom: MTBF jumps from 30 to 120 hours month to month. Root cause: only 3 or 4 failures in the window, so one event shifts the average massively. Fix: pool at least 8 to 10 failure events or several months before trusting a trend, and report a range. With 4 failures the estimate can be off by 50 percent; with 12 the swing tightens considerably. Pair the Downtime Cost per Event tool with failure counts so a single 6-hour outage at 2,300 dollars per hour, 13,800 dollars, is seen as the outlier it is.

Published 2026-07-01.