Your predictive maintenance programme has been running for six months. Sensors are live, alerts are firing and the team is busy. Then finance asks the only question that matters: is it working? If the honest answer is "we think so", you are missing the right numbers. Counting alerts or sensors measures activity, not results. The top predictive maintenance KPIs measure whether failures are being caught early, whether alerts can be trusted and how much money the programme saves. This guide gives you five core KPIs with formulas, a worked cost-avoidance example and a 90-day rollout, and shows how OXMAINT AI, the AI-native CMMS, builds them from the data you already log.
OXMAINT AI · AI-Powered CMMS · Predictive Maintenance KPIs
Top Predictive Maintenance KPIs for Manufacturing Plants in 2027
KPIs fall apart when alerts, inspections and work orders sit in different tools. OXMAINT AI links every step in one maintenance management platform, so catch rate, alert accuracy and cost avoided come from your own records, not spreadsheets.
1Requests & InspectionsQR scans and digital checklists capture what operators and technicians see
➜
2Issues & DefectsA failed check or condition alert is logged against the asset
➜
3Work OrdersEach issue becomes an assigned job with parts and a recorded outcome
➜
4PM & PredictiveResults feed preventive schedules and predictive alerts, then your KPIs
Activity Metrics vs Result Metrics
Many dashboards show how busy the programme is. Leadership wants to know what it delivers. The difference decides whether PdM gets funded again next year. See analytics and reporting in OXMAINT AI.
ACTIVITY METRICS
Shows effort
- Number of sensors installed
- Alerts generated per month
- Assets under monitoring
Useful for rollout. Not proof of value.
RESULT METRICS
Shows impact
- Failures caught before they happened
- Alerts that turned out to be right
- Money saved against a reactive baseline
This is what the five KPIs below measure.
The 5 Predictive Maintenance KPIs — Formula by Formula
Each card gives a plain definition, the formula, what to watch for and a warning. Define each term once, write it down and keep it fixed so trends mean something. Sign up free and start tracking.
1
Catch Rate
Of the failures that happened on monitored assets, how many did PdM warn you about first?
Failures flagged in advance ÷ All failures on monitored assets × 100
Watch: a low rate means blind spots, such as failure modes with no sensor.
2
Alert Accuracy
When PdM raises an alert and someone investigates, how often is there a real problem?
Confirmed alerts ÷ Alerts investigated × 100
Watch: low accuracy breeds alert fatigue, and people stop responding.
3
Detection Lead Time
How much warning do you get between the alert and the point the asset would have failed or was repaired?
Date of repair or failure − Date of first valid alert
Watch: lead time must exceed your time to plan and get parts, or the alert is too late to help.
4
PdM-Generated Work Share
What share of your work orders started from a condition alert rather than a breakdown or a calendar?
Work orders from PdM alerts ÷ Total work orders × 100
Watch: a share near zero suggests alerts are not reaching the work order system.
5
Cost Avoided per Year
The money saved by fixing problems early, minus what the programme costs to run.
(Avoided downtime hours × Cost per hour) + (Reactive repair cost − Planned repair cost) − Programme cost
Watch: count only failures you can show PdM prevented, and agree the assumptions with finance.
What a PdM Scorecard Looks Like
One page, five tiles, reviewed monthly. Green means the programme is doing its job. Amber and red tell you where to look. Book a demo to see KPI dashboards.
Illustrative PdM Scorecard · Example Values
Example figures only. Your targets depend on your assets, failure modes and how long the programme has run.
Worked Example — Cost Avoided, Step by Step
This is the number finance cares about, so show the working. The figures below are examples to show the method, not benchmarks. Start free and use your own costs.
Failures prevented by PdM alerts
6
Downtime avoided: 6 × 8 hrs × $5,000 per hr
$240,000
Repair saving: 6 × ($3,000 difference, reactive vs planned)
$18,000
Programme cost for the year
− $60,000
Net cost avoided
$198,000
Stop Defending the Programme. Start Showing the Numbers.
OXMAINT AI links condition alerts, work orders, parts and downtime so your PdM KPIs come from real records, not memory and spreadsheets.
Supporting KPIs That Complete the Picture
The five core KPIs prove PdM works. These supporting maintenance KPIs show whether the wider programme, including preventive maintenance, is healthy. Explore OEE analytics.
PM Compliance
Preventive tasks done on time ÷ tasks due
Planned vs Reactive Work
Share of work that was scheduled ahead
MTBF
Average run time between failures
MTTR
Average time to restore after a failure
Unplanned Downtime
Hours lost to unexpected stops
OEE Availability
Share of planned time the asset runs
The PdM KPI Maturity Ladder
You do not need all five on day one. Most programmes climb in this order, and each step builds on the last.
LEVEL 1Count alerts and assets monitored
LEVEL 2Track alert accuracy and PdM work share
LEVEL 3Measure catch rate and lead time
LEVEL 4Report cost avoided to finance
4 KPI Traps That Make PdM Look Better Than It Is
!
Counting every alert as a win
An alert is only a win if the problem was real and acted on.
!
Claiming every avoided cost
Credit PdM only where the evidence shows it prevented the failure.
!
Changing definitions mid-year
Move the goalposts and the trend line becomes meaningless.
!
Ignoring misses
Failures PdM did not catch teach you the most. Record them too.
A 30-60-90 Day Plan to Get Your KPIs Live
DAYS 1–30
Define and baseline
- Write each KPI definition
- List monitored assets and failure modes
- Capture last year's downtime and repair costs
DAYS 31–60
Connect and record
- Tag work orders that came from alerts
- Mark each alert confirmed or false
- Log failures and downtime on every asset
DAYS 61–90
Report and improve
- Publish the one-page scorecard
- Review misses and false alerts
- Agree the cost-avoided method with finance
Where Each KPI Comes From in OXMAINT AI
A KPI is only as good as the record behind it. OXMAINT AI keeps those records in one place so the numbers can be traced. Sign up free and see your first dashboard.
PdM Work Share
Work orders generated from condition data, counted against all jobs
Frequently Asked Questions
What are the key performance indicators for predictive maintenance?
Five core KPIs cover it: catch rate, alert accuracy, detection lead time, PdM-generated work share and cost avoided per year.
Sign up free and track them.
What is the difference between predictive and preventive maintenance KPIs?
Preventive KPIs, such as PM compliance, measure whether scheduled work is done. Predictive KPIs measure whether condition alerts are accurate, early and acted on. Track both.
Book a demo to see both views.
How do I calculate the ROI of predictive maintenance?
Add avoided downtime cost and repair savings, subtract the programme cost, and compare with that cost. Use only failures you can show PdM prevented.
Start free and build your case.
How many KPIs should a maintenance team track?
Start with the five core ones plus two or three supporting metrics. A short scorecard that is reviewed monthly beats a long dashboard nobody reads.
Book a demo to pick your set.
What is a good alert accuracy for predictive maintenance?
There is no universal target. Trend it over time and aim for improvement as thresholds are tuned. If accuracy is low enough that people stop trusting alerts, fix that first.
Sign up free and start measuring.
Measure What Predictive Maintenance Actually Delivers.
Bring alerts, work orders, parts and downtime into OXMAINT AI, and walk into every review with five numbers you can defend.
