A single MTBF number tells you the average time between failures — and almost nothing else. Two assets can share an identical MTBF of 10,000 hours while one fails predictably from wear and the other fails early from a manufacturing defect, and a maintenance plan built for one will actively hurt the other. Weibull analysis replaces that single average with a full failure distribution, so an engineer can see whether failures are trending toward infant mortality, random events, or wear-out — and set the interval that actually matches the pattern. See how this changes a PM strategy — book a Weibull analysis walkthrough.
β < 1
Failure rate decreasing — infant mortality, usually a quality or installation issue
β = 1
Constant failure rate — random failures, time-based PM adds little value
β > 1
Failure rate increasing — wear-out, time-based PM is highly effective
63.2%
Share of the population expected to have failed by the characteristic life, η
B10 Life
Industry-standard metric: the time by which 10% of a population is expected to fail
Reliability Engineering Standard
Facility and reliability teams are moving away from fixed, vendor-recommended replacement intervals and toward evidence-based intervals derived from actual failure history. For critical rotating equipment — bearings, pumps, motors, fans — a Weibull-derived B10 life or characteristic life gives a defensible, data-backed number instead of a generic manufacturer default that is often set conservatively for liability, not for your operating conditions.
MTBF-Only Reporting vs. Weibull Distribution Analysis
MTBF is not wrong — it is incomplete. Dividing total operating hours by failure count produces a real average, but averages erase the shape of the underlying failure pattern. Two very different failure stories can produce the exact same MTBF, and only a distribution reveals which one you're actually looking at.
MTBF-Only Reporting
Total operating hours ÷ failure count
IncludedA single average number, useful for rough capacity planning
IncludedSimple to calculate directly from CMMS run-hours and failure counts
IncludedFamiliar terminology most maintenance teams already report on
GapCannot distinguish infant mortality from genuine wear-out
GapTreats an early failure and a late failure as the same event
GapGives one interval for a population regardless of failure pattern
GapNo way to calculate a defensible B10 life or replacement interval
Two assets with an identical MTBF can have completely opposite failure patterns — and opposite correct maintenance strategies
VS
Weibull Distribution Analysis
Shape (β) and characteristic life (η) from failure history
IncludedEverything MTBF reports, plus the actual failure pattern shape
IncludedIdentifies which of the three beta regions a population falls into
IncludedCalculates characteristic life and B10 life for interval-setting
IncludedCorrectly separates still-running suspended units from real failures
IncludedFlags when a failure pattern shifts and a PM strategy needs updating
IncludedRuns directly from CMMS failure history — no standalone reliability tool
A Weibull fit turns the same failure history into a specific, defensible replacement interval — not just an average
The Three Beta Regions Behind Every Failure Pattern
The shape parameter, beta, is what a raw MTBF number cannot show you. Plotting failure rate against time reveals which of three regions an asset population sits in — and each region calls for a completely different maintenance response.
β < 1
Infant Mortality
Failure rate decreases as time passes. Root cause is usually a manufacturing defect, poor installation, or a commissioning error — not age.
Response Fix the root cause and burn-in test new units; time-based PM will not address this pattern.
β = 1
Random Failure
Failure rate stays constant over time. Failures are triggered by external events — power surges, operator error — independent of asset age.
Response Shift to condition monitoring and inspection; a fixed replacement interval adds cost without reducing failures.
β > 1
Wear-Out
Failure rate increases with time — fatigue, corrosion, erosion, and mechanical wear. This is the pattern time-based PM is actually built for.
Response Set a time-based replacement interval near the B10 life, before the failure rate climbs sharply.
Find Out Which Beta Region Your Critical Assets Are In
OxMaint fits Weibull parameters directly from your CMMS failure history, so you know whether a fixed interval will help — or waste money — before you set it.
What Weibull Analysis Needs From Your CMMS
A Weibull fit is only as good as the failure history feeding it. These are the four data inputs a CMMS needs to capture consistently before an engineer can trust the resulting beta and eta values.
01
Failure Event Timestamps
Exact time-to-failure per unit, not just a work order date
Every failure needs a precise operating-time value at the moment it occurred — run hours or calendar time since install or last replacement — not just the date a work order was raised, which can lag the actual failure by days.
02
Suspended (Censored) Unit Data
Units still running, not yet failed
Assets still in service without a failure are just as important as failures themselves. Ignoring these "suspensions" biases the analysis toward shorter lifetimes and produces an artificially pessimistic B10 life.
03
Asset Population Grouping
Same make, model, and operating condition
A valid Weibull fit needs a population of comparable units — same equipment type under similar duty. Mixing a lightly loaded pump with a heavily loaded one of the same model blurs the fit and hides the real pattern.
04
Root Cause & Failure Mode Tags
What actually failed, not just that it failed
Mixing multiple failure modes — a bearing failure and a seal failure — into one dataset produces a distorted beta that represents neither mode accurately. Tagging failure mode at closeout lets each mode be analyzed on its own.
A Worked Example: From Failure Data to a Replacement Interval
A bearing population fitted with Weibull parameters of β = 2.0 and η = 10,000 hours illustrates how the distribution translates directly into an operating decision, not just a chart.
| Operating Hours | Probability of Failure | Reliability | Interpretation |
|---|---|---|---|
| 1,000 hrs | 1.0% | 99.0% | Well within safe operating range |
| 5,000 hrs | 22.1% | 77.9% | Failure risk climbing, plan inspection |
| 10,000 hrs (η) | 63.2% | 36.8% | Characteristic life — most of the population has failed |
| 15,000 hrs | 89.5% | 10.5% | Replacement should already be scheduled well before this point |
Reliability KPIs Worth Tracking Once Weibull Is Live
Fitting a distribution once is a good start; tracking these metrics over time is what turns Weibull analysis into an ongoing reliability program instead of a one-off report.
B10 Life
Interval Basis
Replacement interval tracked against B10 life instead of a fixed vendor default
Δ β
Beta Drift
Change in shape parameter over time — a signal the failure mode itself is shifting
100%
Failure Mode Tagging
Share of closed work orders with a documented failure mode for clean analysis
↓ Trend
Unplanned Failure Rate
Unplanned failures on assets with a Weibull-derived interval, tracked over time
n ≥ 6
Population Sample Size
Minimum comparable units recommended before treating a fit as statistically reliable
$ Saved
Avoided Early Replacement
Cost saved by extending intervals on assets shown to have longer real-world life
Frequently Asked Questions
How much failure history do I need before a Weibull fit is trustworthy?
A handful of failures can produce an initial estimate, but a comparable population of six or more units, including still-running suspensions, gives a meaningfully more stable beta and eta than a fit built on two or three points.
What's the difference between MTBF and the Weibull characteristic life?
MTBF is a straight average of time between failures. The characteristic life, eta, is the time by which 63.2% of the population is expected to have failed — a distribution-based figure that also accounts for the shape of the failure pattern.
Can Weibull analysis run directly from CMMS work order history?
Yes, provided failure timestamps, suspensions, and consistent asset grouping are captured at closeout. Most of that data already exists in a CMMS — the gap is usually in how consistently it gets tagged.
What happens if I set a PM interval without knowing the beta value?
A fixed interval applied to a random-failure (β = 1) population wastes budget with little reliability benefit, while the same interval applied too late on a wear-out (β > 1) population misses failures it could have prevented.
Can I test this against my own equipment history before rolling it out further?
Yes — sign up free and run a fit against one critical asset population, or book a walkthrough to review your failure data together first.
Replace Vendor Defaults With Evidence-Based Intervals
OxMaint turns CMMS failure history into Weibull beta and eta values automatically — so every replacement interval is backed by your actual asset data, not a generic manufacturer recommendation.
