Mean time between failures (MTBF) estimates operating time per failure for repairable equipment. For a stated observation window, divide total operating hours by the number of qualifying failures. Record the hours, failure definition and sample size beside the result. A number without those inputs is difficult to use.
Calculate MTBF
MTBF in hours = total operating hours ÷ qualifying failures.
A packaging machine runs for 720 hours during a reporting period and experiences four functional failures. Its observed MTBF is 720 ÷ 4 = 180 operating hours per failure. This is a historical estimate. It does not mean the machine will fail exactly 180 hours after its next repair.
If the same machine was installed for 1,000 calendar hours but operated for 720, do not substitute 1,000 in the numerator. Nights, idle shifts and planned shutdowns can otherwise make the result look better without changing reliability.
Use the calculator on this page with your own operating hours, repair hours and failure count. Export the result to keep its assumptions with the discussion. No account is needed for the public calculation.
AI-generated editorial illustration; not a customer photograph or product screenshot.
Build a defensible source record
For each asset, retain the reporting window, meter readings or operating logs, failure start, restored time, failure mode and work-order reference. Define whether a failure means loss of the required function, reduced capacity, or another agreed condition. A scheduled inspection is not automatically a failure. Several tickets for the same outage should not become several failures.
A practical weekly review is to inspect the source records before comparing the trend:
- Reconcile meter hours with the operating log. Check resets and replaced meters.
- Link duplicate tickets to one failure event.
- Separate planned maintenance from unplanned loss of function.
- Retain open failures in the event register and explain how incomplete repairs affect each metric.
- Compare the same asset, duty and failure definition across periods.
For a fleet of comparable machines, use total exposure divided by total failures. Two machines with 600 and 900 operating hours and two and three failures give 1,500 ÷ 5 = 300 hours. Averaging individual ratios can mislead when their exposure or failure counts differ. Do not combine unlike assets simply to obtain a smoother headline number.
A minimal calculation record can be kept as CSV:
asset,window,operating_hours,failures,mtbf_hours
PACK-01,example period,720,4,180
Zero failures and small samples
With zero observed failures, the simple ratio has a zero denominator. Report “0 failures in 720 operating hours”, not an infinite or guaranteed MTBF. A short observation window may say little about rare failures. Preserve the exposure and count so a reliability specialist can choose a model and uncertainty interval appropriate to the data.
The NIST reliability handbook distinguishes instantaneous failure rate from average rates. Those concepts should not be collapsed into a claim that a single historical ratio predicts every asset’s next failure.
MTBF, MTTR and availability
MTBF asks how much operating exposure occurred per failure. MTTR asks about time spent restoring or repairing equipment, depending on the declared definition. OEE also considers production speed and good output.
The approximation MTBF ÷ (MTBF + MTTR) is meaningful only when the underlying time definitions and model match. It is not the same as counting how many assets currently have an “operational” status. Nor is it a substitute for measuring capacity lost during actual scheduled production.
AI-generated editorial illustration; not a customer photograph or product screenshot.
What action should follow a low MTBF?
Start with repeated failure modes and the consequences of each event. Compare the replaced component, operating condition, parts specification, procedure and verification. Choose a specific intervention and retain a before-and-after window with comparable exposure. Do not increase every preventive frequency merely because the overall average declined.
A usable investigation record contains the observed problem, supporting work orders, a plausible mechanism, assigned action, due date and a check on recurrence. Changes in load or product mix may explain a trend; a better number alone does not prove that maintenance caused the improvement.
Use the result in PreventiveHQ
Record assets, meter readings, downtime and completed work in the workspace. Use the public calculator when you have verified operating exposure. The reliability report’s elapsed gap between recorded outages is a separate measure and must not be treated as operating-hours MTBF. Manual records are only as complete as the events your team captures.
Try the product tools, install a manufacturing checklist, or inspect current team pricing. Evaluate the product by completing one real maintenance record and retrieving its evidence later.
Specify the asset boundary before collecting failures
An MTBF calculation needs a stable definition of the item being measured. A packaging line can contain conveyors, motors, drives, sensors and a sealing station. A failure of any one component might stop the whole line, but line reliability and motor reliability are different questions. State whether the record describes the complete production function or one repairable component.
For a line-level view, count qualifying interruptions to the required line function. For a component view, retain that component's operating exposure and failure events. Do not divide total line operating hours by an arbitrary collection of component repair tickets and call the result a component MTBF. The numerator and denominator must describe the same boundary.
Document changes to that boundary. If a second conveyor is added, a standby pump is commissioned or a component is replaced with a materially different design, the later observations may describe a different system. Keep the historical record, but explain the change before comparing the periods as if their conditions were identical.
A useful calculation note identifies the asset, included components, excluded events and required function. This does not need to become a lengthy engineering report for every routine review. It needs to be specific enough that another analyst would select substantially the same events and exposure from the underlying records.
AI-generated editorial illustration; not a customer photograph or product screenshot.
Distinguish a failure event from a work order
One outage can generate several work orders: investigation, mechanical repair, electrical support and follow-up inspection. Counting each ticket as a failure inflates the denominator. Conversely, one work order might describe several separate interruptions over a shift. Counting it once could conceal repeated loss of function.
Use an event register or a consistent event reference to connect the work records. Preserve the reported time, affected function, restoration evidence and related jobs. The event should remain identifiable even if a work order is reassigned, split or reopened. Administrative handling should not silently change the historical failure count.
An illustrative machine stops because a drive trips. A technician investigates, an electrician replaces a failed component and a planner creates a later inspection. Those activities can describe one functional failure and several maintenance tasks. If the machine is restored and then suffers a separate qualifying failure the next day, that later event belongs in the count as well.
Agree how to treat repeated stops during an unresolved incident. The right convention depends on the function being measured and the operational record. Write the convention down and apply it consistently. Changing the rule because one reporting period looks unfavorable defeats the purpose of the trend.
Choose the observation window deliberately
A reporting window can follow a calendar month, a production campaign or a defined amount of operating exposure. Each choice answers a slightly different question. Calendar windows are easy to coordinate with management reporting. Campaign windows may better preserve consistent operating conditions. Exposure-based windows can make comparisons easier when utilization changes substantially.
Use the window that supports the decision, then show its dates and exposure. A machine that ran infrequently this month contributes less information about failure behavior than one that operated continuously. A high observed MTBF from a very short, lightly used period should not automatically outrank a longer record with more demanding service.
Keep events near the boundaries explicit. A failure can start before the reporting window and finish inside it, or start inside the window and remain unresolved afterward. The handling of operating exposure and the handling of repair duration are separate accounting choices. Record them so the MTBF trend can be reconciled with the repair-time report.
Do not erase an open event simply because it is inconvenient for a monthly summary. Preserve the event and explain how the reporting rule treats it. The maintenance team may need to act on an unresolved failure even when a particular summary calculation does not yet include a completed repair duration.
AI-generated editorial illustration; not a customer photograph or product screenshot.
Reconcile operating hours with the physical process
An hour meter is useful only when its meaning matches the analysis. Some meters record energized time, some record running time and others derive hours from a control signal. A motor energized while the process is idle may accumulate time differently from a production counter. Confirm what the source measures before treating it as exposure to the failure mechanism under investigation.
Check for counter resets, replacements and unit changes. A lower end reading than start reading is not negative operating exposure. It is an exception requiring investigation. Retain the old reading, the replacement or reset event and the basis for any reconstructed hours. If reliable reconstruction is impossible, disclose the missing interval instead of inventing a smooth series.
For manually entered hours, compare a sample with the original operating logs. Look for copied values, transposed digits and shifts that overlap. The objective is not to demand perfect data before doing any analysis. It is to understand which conclusions the available data can support and where uncertainty is large enough to change the decision.
If calendar time is the only dependable source, report that limitation explicitly and choose a metric name that reflects it. Do not label elapsed calendar hours as operating hours merely to fit an existing dashboard field. A transparent, limited measurement is more useful than a precise-looking ratio built from incompatible inputs.
Work through unequal exposure without averaging ratios
Consider an illustrative pair of comparable pumps. Pump A records 400 operating hours and two failures. Pump B records 1,200 operating hours and three failures. Their individual observed ratios are 200 and 400 hours per failure. The combined exposure is 1,600 hours and the combined failure count is five, giving 320 hours per failure for the defined group.
The simple average of the two individual ratios is 300 hours. That gives each pump's ratio equal weight despite the unequal exposure and failure counts. For the grouped exposure calculation, retain the underlying totals and divide those totals. This also makes it easier to explain which asset contributed which information.
Grouping still requires a defensible boundary. Pumps with different duties, designs or operating environments may be unsuitable for a single performance comparison. A combined number can describe the selected population while concealing meaningful differences within it. Show the asset-level counts beside the total when those differences matter to the maintenance decision.
A pump with zero observed failures is another reason to retain exposure and counts. Its individual simple ratio cannot be calculated with a nonzero denominator, but its valid operating exposure can still contribute to a properly defined grouped observation. Explain the population and method rather than dropping the zero-failure asset from the report.
Use uncertainty to guide the next question
MTBF is an estimate from observations, not an appointment for the next breakdown. A few events can produce a large change in the ratio. If operating exposure remains similar, moving from one observed failure to two can halve the simple estimate without establishing a stable long-term change in failure behavior.
Show the event count prominently and avoid excessive decimal precision. The appearance of a precise number can distract from a weak sample. For an operational discussion, the underlying events, failure modes and conditions often provide more useful information than additional decimal places in the summary.
Formal uncertainty analysis requires assumptions about the process and the data. The NIST section on repairable-system models provides a route into those distinctions. A reliability specialist can help decide whether a model is appropriate when the result supports a consequential engineering or investment decision.
For a routine review, begin with a more modest question: what observations would change the team's interpretation? That might be more operating exposure, better failure classification or a comparison before and after a documented design change. This keeps uncertainty connected to a practical data-collection decision.
AI-generated editorial illustration; not a customer photograph or product screenshot.
Investigate the failure mode before changing the schedule
A lower MTBF does not automatically mean that every preventive task should happen more often. Review the failure modes. Repeated contamination, incorrect installation, an operating change and age-related wear call for different investigations. Increasing inspection frequency without addressing the cause may add work while leaving the underlying problem intact.
Select an event and reconstruct it from the record. Identify the required function, observable loss, affected component, finding and restoration action. Distinguish confirmed findings from early assumptions. A work order that says “bearing problem” may describe a symptom, a diagnosis or an unverified guess; the distinction affects the next maintenance decision.
Then review whether the existing task could reasonably detect or prevent the specific problem. Check the task's applicability, execution evidence, access conditions and response to abnormal findings. A task that repeatedly identifies deterioration but never leads to timely corrective work has a different weakness from a task that measures the wrong condition.
Record the proposed intervention and how it will be evaluated. If the team changes a component, procedure or operating practice, retain the effective date and the affected asset population. The next MTBF review can then examine the observed record in context, without presenting a simple before-and-after comparison as automatic proof of causation.
Build a review record that colleagues can reproduce
A reproducible MTBF record includes the calculation inputs, selection rules and references to the source events. Keep the reporting window, asset boundary, operating-hour source, included failures, exclusions and arithmetic together. A screenshot of the final ratio alone is insufficient when someone later asks why the trend changed.
Assign an owner for resolving ambiguous events. The analyst may understand the calculation while a technician understands what physically failed and an operator understands the lost function. A short review involving those perspectives can prevent a category error from becoming a management target.
Keep the original record when correcting an input. Explain the correction and its effect on the reported result through the organization's supported recordkeeping process. Quietly replacing a failure count makes historical comparisons difficult and can erode confidence in otherwise useful reporting.
Use MTBF alongside other evidence: failure consequences, repair durations, downtime context, operating conditions and maintenance findings. The purpose is to choose a justified next action. A team that can explain one recurring failure and complete the right corrective work has learned more than a team that has merely made its headline ratio larger.
Questions to include when sharing the result
Before sending an MTBF figure to a colleague, check that the note answers five practical questions: which repairable asset or population was measured, how much operating exposure was observed, what qualified as a failure, how many events were included and which observation window was used. These inputs let the recipient interpret the number without guessing at its scope.
Add the decision the result is intended to support. A recurring-failure investigation, a maintenance-plan review and a replacement discussion may use the same underlying events differently. State the known limitations and link the source record so a later reviewer can reproduce the calculation.
If the recipient asks when the next failure will occur, explain that the observed ratio does not supply that date. Use the asset's condition, failure history and appropriate engineering analysis to decide what further assessment is needed. Preserve that boundary even when a dashboard makes the historical average look like a countdown.
AI-generated editorial illustration; not a customer photograph or product screenshot.