An asset criticality analysis ranks your assets by the risk of failure: how severe are the consequences if an installation loses its function, and how likely is that to happen? Each asset ends up with a criticality class, usually A, B or C. With that class you direct maintenance hours, inspections and spare parts to where failure costs the most, instead of giving every pump the same attention.
What an asset criticality analysis is
Criticality is risk, narrowed down to the failure of a single asset. You weigh two things: the consequence if the asset loses its function, and the likelihood that this happens. A machine that breaks down often but does little harm is not critical. A fire pump that almost never fails is, because if it does not start during a fire, the consequence is unacceptable.
That makes criticality analysis an application of ordinary risk management. ISO 31000 describes that process in general terms: identifying, analysing, evaluating and treating risk. A criticality analysis applies it to your plant, and the result is a ranking you can put to work in maintenance straight away. You will also see it called equipment criticality ranking or criticality assessment. The letters A, B and C have nothing to do with the ABC analysis of your stores inventory, which groups items by consumption value.
Criticality per asset or per failure mode
Don't confuse an asset criticality analysis with the criticality in an FMECA. An FMECA scores each failure mode of an asset separately: a bearing that seizes, a seal that leaks, a valve that fails to close. According to IEC 60812, the standard for FMEA and FMECA, the analysis becomes an FMECA once the ranking of failure modes takes at least the severity of consequences into account. Asset criticality works one level up: one score per asset. That makes it quick enough to cover hundreds of assets, which is why you use it to decide which assets deserve an FMECA.
Why record criticality: four decisions that depend on it
No maintenance team has the hours to maintain, inspect and stock spares for every asset with the same care. The criticality class is the allocation key for four decisions:
- Maintenance strategy. Critical assets get a well-founded strategy, often through RCM. For non-critical assets, run-to-failure can be a deliberate and cheap choice.
- Depth of analysis. The class decides which assets go through an FMECA and which can rely on the manufacturer's maintenance advice.
- Spare parts. Whether a part sits on the shelf depends heavily on what a failure costs. How to make that call part by part is covered in which spare parts to keep in stock.
- Work priority. When two failures happen at once, the critical asset goes first, and when the backlog is ranked in maintenance planning and scheduling, the class counts alongside urgency.
The Norwegian standard NORSOK Z-008, on risk-based maintenance and consequence classification, follows the same idea. You break a plant down into main functions and sub-functions, record the redundancy, assess the consequences of losing each function, and use those consequence classes together with likelihood to prioritise work orders and spare parts. The standard was written with the oil and gas industry in mind, but it names manufacturing, processing plants and energy installations among the sectors where its principles apply.
Which consequences to weigh
Score the consequence of failure per category. These five are common:
- Health and safety: could the failure injure someone, or does a safeguard drop out?
- Environment: leaks, emissions, exceeding a permit limit.
- Production or availability: how many hours of output are lost, and can a buffer or stock absorb it? In a waste-to-energy plant this covers not only lost electricity and heat, but also waste that cannot be processed.
- Quality: rejected product, complaints, recalls.
- Repair cost: parts, labour, specialist contractors, express freight.
Only include categories that your organisation actually uses to make decisions; every extra category makes the session longer. Two other factors need their own scoring rules.
Redundancy
If a standby unit takes over the duty, the consequence of one unit failing is smaller. But redundancy only counts if the standby actually starts. A standby pump that sits idle for months can fail without anyone noticing. The reliability data standard ISO 14224 calls this a hidden failure: a failure that is not immediately evident and often only shows up when the function is called on. So test-run standby units periodically or alternate duty and standby, and allow for the periods when one unit is down for maintenance, because the other then stands alone.
A practical rule: first score the consequence as if there were no standby, then lower it by one level if a full-capacity, regularly tested standby takes over straight away. One level, not more, because the standby can fail too. Both units get the same class.
Warning time
Some failures announce themselves weeks ahead through vibration, temperature or oil condition; others come without warning. In an FMECA, detectability is a separate score. At asset level it is better to stick to consequence and likelihood, and to use the warning time when choosing the strategy: an asset with a long, measurable run-up to failure is a good candidate for condition-based maintenance.
How to carry out an asset criticality analysis in 7 steps
- Define the scope and the asset list. Which installation, which boundaries, what level of detail? Work at the level where you plan maintenance and report failures: the pump, not every bolt. A cleaned-up asset register is the starting point; without a clear list you score assets twice or miss them altogether.
- Choose a scoring model. The two common models are a risk matrix (consequence against likelihood) and a points system that adds up scores per category. Also agree how to combine the categories. With an average or weighted sum, a high safety consequence is diluted by low scores elsewhere; with "worst consequence counts", the worst category sets the consequence score. A fire pump scores 5 for safety and 1 for production: the average of the two is 3, the worst consequence is 5.
- Define the scales in concrete terms. "Major production loss" means something different to everyone. Set a measurable threshold for each score, in hours, euros or recognisable situations. For production loss it helps to know what an hour of downtime really costs; see the cost of downtime. Align the thresholds with your own business and, if there is one, with the risk matrix your organisation already uses, so that a high risk means the same everywhere.
- Score in a cross-functional session. Bring together production, maintenance, safety and someone with a view of the failure history. Production knows what a failure costs; technicians know how often things fail and how long repairs take. Back up the likelihood with data where you can: with consistently recorded failure codes you can count how often an asset type actually failed. Write down the assumption behind every score so you can test it later.
- Assign classes. Decide which cells of the matrix belong to which class. In the example below there are three: A, B and C.
- Define what each class means. A class with no effect on the work is just a label. For each class, set the strategy, the inspection intensity, the stocking policy and the priority when it fails.
- Review periodically. A modification, a serious incident or a change in production demand shifts criticality. Schedule a fixed review as well, for example once a year.
The scales from step 3 could look like this. The worked example below uses the same thresholds.
| Score | Consequence: safety and environment | Consequence: production loss | Likelihood of failure |
|---|---|---|---|
| 1 | No injury, no environmental effect | None, or absorbed by a buffer | Less than once in 10 years |
| 2 | Injury without lost time, small spill on site | Less than 4 hours | Once every 3 to 10 years |
| 3 | Lost-time injury, limited emission | 4 to 24 hours | Once every 1 to 3 years |
| 4 | Serious injury, permit limit exceeded | 1 to 7 days | More than once a year, up to 4 times |
| 5 | Permanent injury or fatality, serious environmental damage | More than 7 days | More than 4 times a year |
Worked example: five assets in the risk matrix
The diagram below places five assets from different installations in a five-by-five matrix. The consequence is always the worst category from the table above; the likelihood comes from the failure history. This matrix weighs consequence more heavily than likelihood: an asset with consequence 5 never ends up below orange, and an asset with consequence 1 never above yellow. That is a deliberate choice. A low likelihood is often the result of good maintenance; rank such an asset lower and you cut exactly the maintenance that keeps the likelihood low.
| Asset | Consequence (worst category) | Likelihood | Class |
|---|---|---|---|
| Flue gas fan, line 1 | 5, production | 3 | A (red) |
| Fire pump | 5, safety | 1 | A (orange) |
| Compressor B (backup: A) | 3, production with standby | 2 | B (yellow) |
| Conveyor 3 | 2, production | 4 | B (yellow) |
| Roof fan, hall 2 | 1, no noticeable effect | 3 | C (green) |
Some of the scores explained:
- In this example, each incineration line of the waste-to-energy plant has one flue gas fan (induced draught fan). If it fails, the line comes off load, and repairing an impeller or motor can take more than a week. With a failure once every one to three years, it lands in the red zone.
- The fire pump rarely fails, but here the consequence sets the class.
- Without a standby, compressor B would score consequence 4: compressed air would be gone for days. If B fails, standby compressor A takes over straight away, and A is test-run every month, so consequence 3. Compressor A gets the same class.
- Conveyor 3 in a distribution centre fails most often of the five, but a buffer absorbs most stoppages and repairs rarely take more than a few hours. Frequent failure is not the same as critical, although it is a good reason to look for the cause.
- The roof fan has no noticeable effect when it fails, so it goes into class C. Here, replacing it when it fails is cheaper than preventive maintenance.
Two out of five in class A is a lot, but the example was chosen to show the differences. If most assets in a real plant end up in A, the score thresholds are set too low or the scoring was done on gut feeling.
What each class means for maintenance
| Class | Strategy and inspection | Spare parts | Priority on failure |
|---|---|---|---|
| A | FMECA or RCM, condition monitoring where possible | Critical parts in stock, consider insurance spares | Immediately, including outside working hours |
| B | Interval-based preventive maintenance, periodic inspection | In stock if the lead time exceeds acceptable downtime | Same or next working day |
| C | Corrective as a deliberate choice, check during rounds | Order when needed | In the normal schedule |
Red calls for more than the standard class A treatment. An asset in the red zone carries a risk you do not want to leave as it is: those assets go through an FMECA first, and that should produce a measure such as condition monitoring, a spare motor in stock or a design change. You can see what a risk matrix with assets in the red zone looks like in a maintenance system under reliability and FMECA.
Common mistakes in criticality analysis
- Everything becomes A. Every department defends its own equipment and scores creep upwards until the class no longer distinguishes anything. Agree beforehand that the scales decide, not the negotiation, and check the spread across classes afterwards.
- Scoring on gut feeling. Without defined thresholds and failure history, a score becomes an opinion. Two teams then put the same pump in different classes.
- Averages instead of the worst consequence. A safety risk then disappears into the middle of the range.
- Redundancy that only exists on paper. A standby that never runs is not a standby.
- Confusing condition with criticality. An old asset in poor condition may have a high likelihood of failure, but it is only critical if the consequence is high too. Condition belongs with likelihood, not with consequence.
- No follow-up. If the class changes nothing about strategy, stock or priority, all you have made is a spreadsheet.
- A one-off exercise. Plants change. An analysis from five years ago no longer holds after two extensions and a shift in production demand.
Frequently asked questions
How many criticality classes should you use?
For most maintenance teams, three classes, A, B and C, are enough. Each class needs to be handled differently in practice, in strategy, stock and priority. With five or more classes that distinction is often no longer made, and the analysis looks more precise than it is. If your organisation already has a company-wide risk matrix, align with its categories so that red means the same thing everywhere.
Who should take part in a criticality analysis?
At a minimum: someone from production or operations, an experienced technician or maintenance planner, someone from health, safety and environment, and the reliability or maintenance engineer who leads the session. Production knows what a failure costs; technicians know how equipment fails. Keep the group small and prepare the asset list and failure history in advance, so the session is spent scoring rather than searching for data.
How often should asset criticality be reviewed?
Review criticality after any change that affects consequence or likelihood: a modification, a new standby unit, a serious incident or a change in production demand. A fixed review is sensible too, for example annually or after a major shutdown. If you recorded the assumption behind each score, you can quickly see at the review which scores no longer hold.
Can you calculate criticality from failure data?
Partly. You can support the likelihood with failure data, such as the number of failures per year or the MTBF of similar assets. Not the consequence: that follows from the asset's role in the process, the redundancy and what downtime costs. If there is no history, score the likelihood from experience and manufacturer data, and refine it with your own data after a year or two.
Which assets should you start with?
Start with one installation or production line where downtime is expensive or safety risks are highest, and score every asset there. You get to know the model, find out which thresholds don't work and quickly have a usable ranking. Then extend line by line. Teams that try to score the whole site at once often lose support before anything has been done with the outcome.