Failure Rate Calculator
Failure rate answers a plain question: for every hour your units run, how many failures should you expect? Enter how many units you tested or ran, how many of them failed, and how many hours each unit operated, and this calculator returns the failure rate in failures per hour. It does one job: it builds the total operating time from your units and hours, then divides. It does not display MTBF, availability or a reliability curve, and the sections below show how to read those from the rate by hand where they apply.
Quick Answer
Testing 200 units for 1,000 hours each with 5 failures gives a failure rate of 0.000025 failures per hour: 5 ÷ (200 × 1,000) = 5 ÷ 200,000. Read on the scale engineers actually quote, that is 25 failures per million unit-hours, and its reciprocal, 1 ÷ 0.000025 = 40,000 hours, is the mean time between failures for the population under a constant-rate assumption.
Key takeaways
- The denominator is total operating time, units × hours per unit, not the calendar time the test took.
- The result is shown to six decimal places. Small real rates can round to 0.000000 on screen, so rescale per million hours before concluding anything.
- MTBF is the reciprocal of the rate, and only under a constant-rate assumption. This tool shows the rate; the reciprocal is a hand step, shown in the worked example.
- A zero-failure window returns 0.000000. That means none observed in that window, never that the product cannot fail.
- Quick Answer
- How to use the failure rate calculator
- The failure rate formula
- Worked example: 200 relays, 5 failures, 1,000 hours each
- Reading the result: per hour, per million hours, FIT and MTBF
- When one rate is honest: the bathtub curve and the constant-rate assumption
- What competing calculators miss
- Frequently asked questions
- Add This Calculator to Your Website
- References
How to use the failure rate calculator
Enter the three counts from one test or observation window, then click Calculate. The results stay hidden until you do.
- Total number of units: every unit that operated in the window, failed and unfailed alike. It must be greater than zero.
- Number of failed units: how many of those units failed in the same window. It cannot exceed the total, and the tool rejects it if it does.
- Operating time per unit (hours): the hours each unit ran. The tool assumes a common figure per unit, which matches standard life-test setups where every sample runs the same duration.
- Read the rate: failures per hour to six decimal places, with the total operating time it was divided by and the same rate rescaled per million unit-hours.
If your units ran for unequal times, the formula still works with one adjustment: total operating time is what matters, so enter the unit count with the average hours per unit (total unit-hours ÷ units), or equivalently set units to 1 and hours to the full total unit-hours. Both give the identical rate, because the tool only ever uses the product. Counts should be whole numbers in real data, but be aware the tool reads entries as decimal numbers and will accept a fractional count without complaint, so sanity-check what you typed.
The failure rate formula
Failure rate λ = Number of failures ÷ Total operating time, and in this tool total operating time = Units × Hours per unit. One division, with the denominator built from your test population.
Total operating time = Total units × Operating time per unit
MTBF (constant rate, by hand) = 1 ÷ λ = Total operating time ÷ Failed units
The unit question deserves care, because this is where published explanations often blur. The time unit inside the rate is whatever unit you enter: the field here is labelled hours, so the result is failures per hour, and a 1,000 in that field means 1,000 hours even if your test log recorded days. Enter days and you get failures per day instead, with no conversion applied. The NIST/SEMATECH e-Handbook of Statistical Methods frames the same quantity more formally in its product reliability chapter: failure rate and hazard rate describe how a population of non-repairable items fails over time, which is the population this calculator's units-times-hours structure describes. The IEEE Technology Navigator's failure rate overview collects the engineering literature behind the same definitions, including the Weibull and bathtub-curve models covered below.
Worked example: 200 relays, 5 failures, 1,000 hours each
A manufacturer life-tests 200 electrical relays. Each relay runs 1,000 hours and 5 relays fail during the test. These are the values pre-filled in the calculator.
| Check (units / failed / hours each) | Result shown | Note |
|---|---|---|
| 100 / 2 / 500 | 0.000040 failures/hour | 50,000 unit-hours, 40 per million |
| 10 / 10 / 100 | 0.010000 failures/hour | Every unit failed: the highest rate this shape of test can return at 100 hours |
| 50 / 0 / 1,000 | 0.000000 failures/hour | Genuine zero: no failures observed in 50,000 unit-hours |
| 1,000 / 3 / 8,760 | 0.000000 failures/hour | Rounding case: the true rate is about 0.00000034 per hour (0.34 per million) |
| 250 / 12 / 2,000.25 | 0.000024 failures/hour | Fractional hours are accepted as entered |
Each row is a real run of this calculator. Read the worked example one step further by hand: the MTBF reciprocal is 1 ÷ 0.000025 = 40,000 hours, which is also 200,000 unit-hours ÷ 5 failures. Both routes agree because they are the same division inverted, and both rest on treating the rate as constant across the window. The per-million rescaling, 25 failures per million unit-hours, is the figure to compare against datasheets, which rarely quote a decimal as small as 0.000025 directly.
Reading the result: per hour, per million hours, FIT and MTBF
A per-hour rate is the correct unit and an awkward one, because reliable products produce very small decimals. Three rescalings make the same number usable, and all three are readings of the single result, not new measurements.
The six-decimal display sets a floor on what the screen can tell you. Any true rate below 0.0000005 per hour rounds to 0.000000, as the 1,000-unit year-long check in the table above shows: 3 failures over 8,760,000 unit-hours is a real, nonzero rate of about 0.34 per million unit-hours that the headline display cannot distinguish from a perfect zero. The badge and the per-million reading in the results box use the unrounded rate, so they still separate the two cases. When a result shows 0.000000 with failures in your inputs, trust the rescaled reading and lengthen the test or pool more units to pin the rate down.
On MTBF, precision about terms matters. Strictly, MTBF (mean time between failures) belongs to repairable systems that fail, get repaired and fail again, while a one-shot test population like this one gives a mean time to failure. Field practice uses MTBF for both, and the reciprocal arithmetic is identical, which is why this page shows the reciprocal openly as a reading rather than pretending the tool computes a separate MTBF output. For neighbouring quality measures with different denominators, the defect rate calculator counts defective items per items produced with no time dimension, and the defect density calculator normalises defects by size, the software-quality cousin of this rate.
When one rate is honest: the bathtub curve and the constant-rate assumption
A single failure rate summarises a window truthfully only when the underlying rate is roughly constant inside it. Real populations follow the bathtub pattern the NIST/SEMATECH reliability chapter models: early failures from defects, a long middle of random failures, then wear-out. One averaged number per phase is defensible; one number across all three is not.
| Phase | Rate behaviour | What one averaged rate means there |
|---|---|---|
| Infant mortality | Falling: weak units fail early | An average overstates later risk. Burn-in and screening are the fixes, not a better average. |
| Useful life | Roughly constant: random, stress-driven failures | The home ground of this calculator. The rate and its MTBF reciprocal are meaningful planning figures. |
| Wear-out | Rising: fatigue, corrosion, depletion | An average understates late risk. Scheduled replacement beats rate-watching here. |
In Weibull terms the same point is one parameter: a shape parameter below 1 fits the falling phase, exactly 1 the constant phase, and above 1 the rising phase, as the TIBCO documentation on the Weibull distribution sets out. Practitioners who track how event counts accumulate against exposure over time, rather than in one test window, are solving the neighbouring problem covered by the exposure rate calculator, and actuaries counting expected events in a period use the expected frequency calculator. Keep the windows separate in your own data too: a rate from a 100-hour bench test and a rate from a year of field returns describe different phases and should never be averaged together, a point reliability practitioners such as Accendo Reliability's survey of the six failure patterns make with field data.
What competing calculators miss
Most failure rate tools ask for two numbers, failures and a total operating time you must build yourself. That design choice drives most of the differences:
- Two-input tools (Tractian, CalculatorsHub, Cryotos, BraveCalculator, TrueGeometry): failures ÷ total operating time directly. Clean if you already track total unit-hours; unhelpful when your raw data is a test population, which is how life-test and warranty data actually arrives. CalculatorsHub adds the MTBF inverse and an exponential-model route, and its MTBF reference table is genuinely useful, but the total time is still your homework.
- Reliability suites (InventiveHQ, FIRGELLI): MTBF, MTTR, availability, reliability over time and downtime in one calculator. Broader and well explained, but the primary input is a single system's operating hours, a repairable-system framing, and the plain test-population rate gets buried inside a suite.
- Guide-led pages: sound bathtub and Weibull explanations with the arithmetic in prose. Fine reading, nothing to rerun with your own counts.
- The gap this page fills: the test-population structure itself. Units × hours per unit is the denominator as recorded, the failed-units guard (failures cannot exceed units tested) is a validation two-input tools cannot even express, and the display limits are stated where they bite: six decimals, the 0.000000 rounding case shown with a real run, and MTBF labelled as a reciprocal reading rather than a phantom second output.
Frequently asked questions
Failure rate, written λ (lambda), is the number of failures per unit of operating time. This calculator builds it as failed units divided by total operating time, where total operating time is the number of units multiplied by the hours each unit ran. A rate of 0.000025 failures per hour means about 25 failures per million unit-hours of operation.
MTBF is the reciprocal of the failure rate under a constant-rate assumption: MTBF = 1 ÷ λ, which is the same as total operating time divided by the number of failures. This calculator displays the failure rate only. For the worked example, 1 ÷ 0.000025 = 40,000 hours, computed by hand from the rate the tool returns.
Because the result is rounded to six decimal places. A true rate below 0.0000005 per hour displays as 0.000000: 3 failures across 1,000 units run for 8,760 hours each is a real rate of about 0.00000034 per hour. Rescale before judging it: multiply the unrounded rate by 1,000,000 to get failures per million unit-hours, which is what the results box reading does.
Yes, whenever no unit failed during the window: the tool returns 0.000000 failures per hour. Read it as no failures observed in that much operating time, not as proof the product cannot fail. The MTBF reciprocal is also undefined at zero failures, since dividing by a zero rate has no finite answer; a longer window or more units is the way to bound the true rate.
An unplanned loss of function that takes the unit out of service until it is repaired or replaced. Planned maintenance stops, minor interruptions that do not stop operation, and failures caused by using the unit outside its specified conditions are normally excluded. Decide the definition before the test starts and apply it to every unit the same way, or the rate is not comparable between tests.
Add the individual running times into one total of unit-hours, then enter the unit count with the average hours per unit (total unit-hours divided by units). The tool only uses the product of units and hours, so the average gives the identical rate. Entering 1 unit with the full total as its hours works the same way.
Usually not. Populations typically show a falling rate early on as weak units fail (infant mortality), a long roughly constant stretch of random failures (useful life), and a rising rate at the end as wear-out mechanisms dominate. A single rate from this calculator is an average over your window and describes the useful-life phase best; compute separate rates per phase rather than one across a whole lifetime.
FIT is failures in time, the electronics industry's unit of one failure per billion unit-hours. Multiply the per-hour rate by 1,000,000,000: the worked example rate of 0.000025 per hour is 25,000 FIT. The per-million-hours figure here is the intermediate scale, rate multiplied by 1,000,000, giving 25 for the same example.
Add This Calculator to Your Website
Embed this calculator on your own site with the code below. Pick a size to preview it live, then copy the code.
Powered by Calculator Zilo
References
- National Institute of Standards and Technology. NIST/SEMATECH e-Handbook of Statistical Methods, 8.1.2.1: Repairable systems, non-repairable populations and lifetime distribution models. Accessed October 21, 2026.
- National Institute of Standards and Technology. NIST/SEMATECH e-Handbook of Statistical Methods, Chapter 8: Assessing Product Reliability. Accessed October 21, 2026.
- IEEE Technology Navigator. Failure Rate. Accessed October 21, 2026.
- TIBCO Software. The Weibull Distribution: Weibull and Reliability/Failure Time Analysis. Accessed October 21, 2026.
- Accendo Reliability (Fred Schenkelberg). Living With The 6 Failure Patterns. Accessed October 21, 2026.
- Wikipedia. Failure rate. Accessed October 21, 2026.