MTBF

Running MTBF as a Monthly Reliability Management System

MTBF times cost per event is the annual price of unreliability, asset by asset. How to count failures honestly and run the number as a monthly system.

MTBF is the denominator of your maintenance budget whether you calculate it or not. Mean time between failures times downtime cost per event determines what unreliability costs per year, asset by asset, and that product is the money case for every reliability investment you will ever propose. A pump with a 730 hour MTBF running 8,000 hours a year fails roughly 11 times annually; at 4,000 dollars per event that is 44,000 dollars a year from one asset. Double the MTBF and you bank 22,000 dollars, every year. Plants that track MTBF by asset class run maintenance as an investment portfolio; plants that do not run it as a complaint department.

The arithmetic is operating hours divided by failure count, and the discipline is in the inputs. An asset that ran 4,380 hours last year, actual operating hours, not calendar hours, and failed 6 times has an MTBF of 730 hours. For a fleet, sum operating hours across identical units: ten pumps running 7,500 hours each with 25 total failures gives 75,000 divided by 25, or 3,000 hours fleet MTBF. The MTBF calculator handles the division; your work is the counting. Use runtime meters or PLC data for hours, because an asset operating 60 percent of calendar time will show a flattering, false MTBF if you count the calendar.

Definitions decide whether the metric means anything. Write down what counts as a failure and hold the line: an unplanned functional loss requiring intervention is a failure; a planned component replacement on condition is not; a 5 minute reset may or may not be, but decide once and count consistently. Mixing planned maintenance into the failure count can understate MTBF 30 to 50 percent and punishes exactly the proactive behavior you want. Track MTTR separately; a plant can double MTBF while repairs get slower and never notice if the two are blended into availability alone. One page of definitions, agreed between maintenance and operations, protects years of data.

Benchmarks vary enormously by asset class, so never quote a universal number. ANSI process pumps average 12 to 18 months MTBF across industry, while top decile refineries run the same pumps past 60 months, a fourfold spread on identical hardware that is entirely installation, operation, and maintenance practice. Rolling element bearings, properly specified, lubricated, and aligned, should reach their L10 design life; in average plants most never do. The practical benchmark is your own trend: a rolling 12 month MTBF by asset class, improving 15 to 25 percent per year during an active reliability program, flat or declining when the program is theater.

The levers are precision, lubrication, and operating context. Precision alignment and balancing matter more than most budgets admit: misalignment of a few thousandths cuts bearing and seal life 30 to 50 percent, and laser alignment on every rebuild is one of the highest return practices in maintenance. Lubrication discipline is next: contamination and wrong or degraded lubricant drive roughly half of bearing failures, and a 3,000 dollar filtration and single point lubricator program on critical assets routinely doubles component life. Operating context closes the set: a centrifugal pump run 30 percent away from best efficiency point can see seal and bearing life fall by half. Fix how machines are installed and run, and MTBF follows.

Respect the statistics at low failure counts. An asset with 2 failures in a year has an MTBF estimate with a confidence interval spanning roughly threefold in either direction; celebrating a 40 percent improvement on 3 data points is noise worship. Aggregate to asset class or fleet level where counts reach 15 to 20 failures before drawing trend conclusions, and use rolling 12 month windows rather than monthly snapshots, which whipsaw. For critical single assets with rare failures, supplement MTBF with leading indicators: PM compliance, vibration alarm counts, lubrication route completion. MTBF tells you where you have been; the leading indicators tell you where you are going.

Run MTBF as a monthly system. Monthly, refresh rolling 12 month MTBF by asset class and publish a top 10 bad actor list ranked by failure count times cost per event; that list is the RCA queue and the reliability budget, in order. Weekly, review new failures against the definitions so the counts stay clean while memories are fresh. Quarterly, pick the two worst actors, complete formal RCAs, fund the fixes, and log the expected MTBF gain. Annually, compare each asset class trend against the prior year and against the cost of unreliability in dollars; kill any program element that moved neither number in 12 months.

World class reliability shows up in the numbers together: MTBF by asset class doubling over 2 to 3 years, proactive work above 75 percent of maintenance hours, PM compliance above 95 percent, and a bad actor list where the names change because problems get solved. The cost signature follows: maintenance cost as a percent of asset replacement value drops toward 2 percent while availability climbs, because reliability is cheaper than repair at any volume. None of it starts with software. It starts with runtime hours, an honest failure count, a one page definition sheet, and a monthly meeting that treats mean time between failures as money, which it is.

Published 2026-07-02.