Key summary points
- Most condition monitoring programmes generate more alerts than teams can meaningfully act on. The problem is not the technology: it is the operating model.
- Treating all alerts equally creates two failure modes: chasing everything and burning out, or ignoring everything and missing the warning that matters.
- A three-tier structure (Watch, Plan, Act now) gives teams a clear decision framework without adding bureaucracy.
- Actionable alerts need three things: evidence, context, and a named owner. Alerts without all three are just notifications.
- Connecting your monitoring platform to your CMMS closes the loop between detection and planned work, and removes the manual admin that causes delays.
- Prediction without action is theatre. The value of condition monitoring is only realised when a detected problem becomes a completed work order.
When the alerts never stop, nothing gets done
Most maintenance teams are not short of data. They are short of signal in all the noise.
If you have deployed condition monitoring on your critical assets, you already know the problem. The platform fires alerts. Emails land. Phones buzz. And somewhere inside that wall of notifications is a genuine warning that needs acting on today. The rest can wait, or should never have been sent in the first place.
The result is predictable. Teams start ignoring alerts. Acknowledgements pile up without follow-through. The monitoring system becomes a background hum rather than a decision-making tool. And when something does fail, the post-mortem usually finds that the warning was there. It just got lost.
This is not a technology problem. It is an operating model problem.
Prediction without action is theatre. The goal of any condition monitoring investment is not to generate more data: it is to generate fewer surprises. To do that, you need a clear, reliable bridge between what the system detects and what the maintenance team actually does. This post sets out how to build that bridge.
Why alerts become noise
The problem usually starts with good intentions. When a monitoring system is first deployed, every alert feels important. Thresholds get set conservatively. Notifications go to everyone. The team is engaged and responsive.
Then volume increases. More assets are connected. Thresholds that made sense in week one start firing regularly on equipment that is running fine. The ratio of meaningful alerts to background noise shifts, and the team adjusts accordingly. They stop treating every notification as urgent because experience has taught them that most are not.
This is entirely rational behaviour. But it creates a dangerous blind spot, and the data tells us the stakes are high.
Two-thirds of companies experience unplanned downtime at least once a month, at an average cost of $125,000 per hour.
ABB Value of Reliability Report [1]
Unplanned downtime now accounts for 11% of annual revenues among the world’s 500 largest companies, equivalent to $1.4 trillion per year.
Siemens True Cost of Downtime 2024 [2]
Industry research also shows that the accuracy of many predictive maintenance alert systems is lower than 50%, meaning maintenance teams regularly arrive at assets to find nothing wrong. That pattern of false positives erodes trust and pushes teams towards ignoring alerts altogether. Once that culture sets in, it is very difficult to reverse.
The accuracy of many predictive maintenance solutions is lower than 50%, creating headaches for maintenance organisations that run to an asset only to find it operating normally.
IoT Analytics Predictive Maintenance Market Report [3]
The issue is that not all alerts carry the same weight. Treating them all the same way pushes teams toward one of two failure modes: either everything gets chased regardless of priority, which burns resource and credibility quickly, or everything gets ignored, which defeats the purpose of having a monitoring system at all. Neither is acceptable. The fix is to stop treating alerts as a single category and start treating them as a tiered system with different rules for each level.
Three tiers that actually work
A practical alert structure does not need to be complex. Three tiers, clearly defined, with agreed roles and response rules, is enough.
Watch. This is a condition worth tracking but not yet a reason to act. Something has shifted from normal baseline: perhaps a gradual rise in vibration on a motor, or a slow drift in temperature. Nothing is broken. Nothing is imminent. But the trend needs to stay on the radar. The right action is to log it, set a review date, and check it again at the next scheduled opportunity. No emergency. No work order. Just awareness.
Plan. This is a condition that needs attention before it becomes a problem. The signal has crossed a threshold that indicates genuine degradation. The asset is not going to fail this afternoon, but it will not make it through the next few months without intervention. The right action is to raise a planned work order in the CMMS, schedule it within the next maintenance window, and make sure the right parts and skills are available. This is where prediction earns its value: turning a future failure into a scheduled job.
Act now. This is a condition that requires immediate response. Risk to production, safety, or quality is real and present. This alert goes to a named person with authority to act, not to a distribution list. Response time is measured in hours, not days. If the situation cannot be resolved in the current shift, the operations manager needs to know.
The discipline here is in the definitions. ‘Plan’ does not mean ‘sometime in the next quarter.’ ‘Watch’ does not mean ‘ignore until it gets worse.’ Each tier needs a clear threshold, a clear owner, and a clear response time. Write it down, agree it with the team, and review it quarterly.
What makes an alert actionable
A tiered structure only works if the alerts within it are actually useful. That means three things: evidence, context, and ownership.
Evidence is the data that justifies the alert. An actionable alert does not just say ‘high vibration on pump 4.’ It says what the reading is, how it compares to the asset’s own historical baseline, how long the condition has persisted, and what failure mode the reading is consistent with. AssetMinder’s health scoring approach does this by weighting alerts according to severity and rolling them up into a single asset health score, so the team can see at a glance whether an alert is an outlier or part of a deteriorating trend.
Context is everything that sits around the data. Has this asset been through a recent repair? Is it running at higher load than normal due to a production surge? Is this a known seasonal behaviour? A platform that surfaces historical trend data alongside the live alert gives the maintenance manager the information needed to make a confident judgement call. Without context, even a well-evidenced alert can be misread.
Ownership is the most commonly overlooked element. An alert that goes to a team inbox or a group chat is an alert that belongs to nobody. Actionable alerts need a named owner: a person who is responsible for reviewing it, deciding what tier it sits in, and either raising a work order or documenting why they have not. If that sounds like extra admin, consider the alternative: an inbox full of acknowledged alerts and a failure that nobody saw coming.
Industry analysis reinforces why this matters. False positives are not just an inconvenience: they are a cultural setback that dims confidence in digital transformation. When maintenance teams learn through experience that most alerts do not result in confirmed faults, they stop responding with urgency, and that behavioural shift is very difficult to reverse.
Predictive maintenance can reduce maintenance costs by up to 25% and increase uptime by 10-20%, but only when the programme converts detections reliably into action.
Deloitte, cited in MaintainX State of Industrial Maintenance 2025 [4]
Connecting alerts to your CMMS
The final link in the chain is integration. Getting an alert right and then manually transcribing it into a work order in a separate system is an opportunity for error, delay, and friction. It also means the alert lives in one place and the work record lives in another, making it very difficult to measure whether your monitoring programme is actually reducing failures over time.
The principle is simple: the system that detects the problem should be able to talk directly to the system that manages the work. In practice, that means a connection between your condition monitoring platform and your CMMS, whether via a direct integration, an API, or a structured export that maps to your work order format.
When this connection works well, a ‘Warning’ alert in AssetMinder can trigger a draft work order in your CMMS with the asset details, the fault description, the evidence trail, and a suggested priority already populated. The maintenance planner reviews it, assigns it, and schedules the job. No double-keying. No information lost in translation. The work record sits alongside the sensor data that justified it.
For sustainability and compliance purposes, this closed-loop record matters too. A clear audit trail connecting a detected condition to a planned intervention to a completed work order gives you auditable evidence that your maintenance programme is proactive and risk-based. That is increasingly relevant for ESG reporting, insurance assessments, and regulatory conversations.
The integration does not need to be perfect from day one. Start with a structured export. Build the habit of matching alert records to work orders. The discipline of the process matters more than the technology at the outset.
Fortune 500 companies are estimated to save 2.1 million hours of downtime and $233 billion in maintenance costs annually through full adoption of condition monitoring and predictive maintenance
Siemens True Cost of Downtime 2024, cited in MaintainX [4]
What good looks like in practice
A maintenance team running this model well operates in a steady state rather than a permanent crisis. The watch list is reviewed weekly. Plan-tier items are on the schedule before they become urgent. Act-now alerts are rare, because the watch-and-plan process is catching most things early.
Operations managers stop being surprised by last-minute shutdowns for ‘unexpected’ failures. Because the maintenance team is working off a shared picture of asset health, conversations about planned downtime windows are based on evidence rather than guesswork. The negotiation shifts from ‘we need to take this line down now’ to ‘we need a four-hour window in the next two weeks.’
The business case becomes visible faster too. When work gets done as a result of a predicted condition, that outcome is recorded. Over time, the pattern of ‘detected, planned, fixed before failure’ builds a convincing body of evidence for continued investment in condition monitoring.
AssetMinder customers have seen this in practice, with the platform contributing to over 7,000 potential failures avoided and more than 5,000 hours of unplanned downtime recovered across monitored sites. Those numbers only materialise when prediction connects reliably to planned action.
Predictive maintenance programmes that close the loop between detection and execution reduce downtime by between 35% and 50%, and extend asset lifespan by between 20% and 40%.
McKinsey, cited in Predictive Maintenance Software Statistics [5]
Three things to take away
- Tiering stops ‘everything is urgent’ behaviour. Without tiers, every alert competes for the same attention. With tiers, the team knows exactly which category of problem they are dealing with and what response is expected.
- Actionable alerts need evidence, context, and ownership. A well-evidenced alert with historical trend data, contextual notes, and a named owner is a decision-making tool. An alert without those elements is just a notification.
- Integrations should remove admin, not add it. The test is simple: does a detected condition flow into a planned work order with less effort than doing it manually? If not, the process needs reviewing.
Frequently asked questions
Alert fatigue sets in when thresholds are set too conservatively or when all alerts are treated with equal urgency. As monitoring scales across more assets, the volume of notifications grows. Teams that cannot distinguish between a minor drift and a genuine warning start ignoring both. The solution is a tiered alert structure that assigns different response rules to different severity levels, so teams know immediately how to respond without having to make a fresh judgement call on every notification.
A Watch alert signals that something has changed from the asset’s normal baseline but does not yet require intervention. The appropriate response is to log it and set a review date. A Plan alert indicates genuine degradation that will require maintenance before the next scheduled opportunity. The appropriate response is to raise a planned work order in the CMMS. The key distinction is urgency: Watch is awareness, Plan is action within a defined window.
An actionable alert needs three elements. First, evidence: the reading, how it compares to the asset’s historical baseline, and what failure mode it suggests. Second, context: recent repairs, current load conditions, and any known operational factors that might affect the reading. Third, ownership: a named person who is responsible for reviewing the alert and deciding what to do next. Alerts without all three elements are notifications, not decisions.
Not necessarily. While a direct API integration is the ideal long-term solution, you can start with a structured data export that maps to your CMMS work order format. The priority is building the habit of linking alert records to work orders, so that detected conditions lead reliably to planned jobs. The discipline of the process is more valuable than the sophistication of the technology in the early stages.
When alerts are tiered and connected to planned work orders, operations managers gain advance visibility of upcoming maintenance needs rather than receiving last-minute disruption notices. Planned downtime windows can be negotiated with production schedules in advance, rather than being imposed by emergency breakdowns. The conversation shifts from reactive to collaborative, which benefits both functions.
A closed-loop record that connects a sensor detection to a planned work order to a completed job provides auditable evidence that your maintenance programme is proactive and risk-based. This kind of documentation is increasingly valuable for ESG reporting frameworks, insurance risk assessments, and regulatory conversations. It also supports the case that condition-based maintenance reduces energy waste and unnecessary parts consumption compared to time-based servicing.
The framework itself can be defined and agreed within a single working session. The practical steps are: classify your current alerts into three tiers, assign response rules and owners to each tier, and document the thresholds that trigger movement between tiers. The initial configuration should be reviewed after 90 days to refine thresholds based on what the team has learned. The technology does not need to change for the operating model to improve.
References
The following sources are cited in the text above. All data and statistics are drawn from publicly available research and industry reports.
[1] ABB Value of Reliability Report: Industrial downtime costs up to $500,000 per hour
[2] Siemens True Cost of Downtime 2024 (via ISM World / IFactory App)
[3] IoT Analytics: Predictive Maintenance Market 5 Highlights 2024
[4] MaintainX: 25 Maintenance Stats, Trends, and Insights for 2026
[5] McKinsey predictive maintenance impact data, cited in Predictive Maintenance Software Statistics
Like this post?
Then check out our last post around building the business case of condition monitoring: Asset health is not a maintenance topic. It is business continuity.
