--- title: "How Utilities Measure Grid Reliability: SAIDI, SAIFI, CAIDI and Related Metrics" canonical: https://aboutshrey.com/post/grid-reliability-metrics/ date: 2026-10-12 10:13:12 description: "Learn how utilities measure grid reliability using SAIDI, SAIFI, CAIDI and related metrics, including formulas, examples, limitations and utility applications." keywords: "grid reliability metrics" categories: [Grid Reliability, Reliability & Outages] tags: [AI and machine learning, AMI, electric grid, electric utilities, electrical engineering, grid modernization, power infrastructure, power systems, smart grid, utility engineering] --- Reading Time: 14 minutes Grid Reliability · Reliability & Outages Category:[Grid Reliability](https://aboutshrey.com/post/category/grid-reliability/)[Reliability & Outages](https://aboutshrey.com/post/category/grid-reliability/reliability-outages/)By Author[Author Shrey](https://aboutshrey.com/post/author/websmaniacinc/)Approx. 12 min readShare:[LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Faboutshrey.com%2Fpost%2Fgrid-reliability-metrics%2F)[Facebook](https://www.facebook.com/sharer/sharer.php?u=https%3A%2F%2Faboutshrey.com%2Fpost%2Fgrid-reliability-metrics%2F)[X](https://twitter.com/intent/tweet?url=https%3A%2F%2Faboutshrey.com%2Fpost%2Fgrid-reliability-metrics%2F) Utilities use grid reliability metrics to quantify how often customers experience interruptions and how long those interruptions last. The most common indices are SAIDI (duration), SAIFI (frequency) and CAIDI (average interruption duration), supported by MAIFI for momentary interruptions and ASAI for service availability. Each turns outage records into a number that can be tracked, compared and used for planning. No single metric tells the whole story. This guide goes deeper into measurement: what each metric measures, how it is calculated, how engineers interpret it, and what it cannot show. For definitions and drivers, see [electric grid reliability](https://aboutshrey.com/electric-grid-reliability/), and for the relationship with disruption recovery, see [grid reliability vs resilience](https://aboutshrey.com/grid-reliability-vs-resilience/). Reliability Measurement → What Each Metric Measures → How It Is Calculated → How It Is Interpreted → What It Cannot Tell Engineers → How Reliability Data Supports Operations and Planning **Key Takeaway** - Grid reliability metrics convert outage and customer-interruption data into measurable indicators of interruption frequency, duration and service performance. - SAIDI, SAIFI and CAIDI provide complementary views, not one complete measure of reliability. - Reporting methodology, data quality and context determine how a number should be read. ## What Are Grid Reliability Metrics? Grid reliability metrics are standardized indices that quantify the frequency and duration of customer interruptions and the availability of electric service. Utilities measure reliability to track performance, report to regulators, find weak areas and justify improvements. The raw inputs are customer interruption data: the start and end of each event, its duration and the number of customers affected. The metrics summarize these into system performance indicators used in reporting, planning and improvement. Different metrics measure different dimensions. No single reliability metric fully describes grid reliability. ## How Utilities Collect Reliability Data Grid Event → Protection / Field Device → Meter / Sensor → SCADA / AMI / OMS → Event Validation → Customer Interruption Record → Reliability Database → Metric Calculation → Reporting → Planning / Improvement - **OMS:** records outages, affected customers, times and causes. - **SCADA:** reports device operations and alarms. - **AMI and smart meters:** provide customer-level power-loss and restoration signals. - **Customer calls and field reports:** confirm and add detail. - **GIS:** maps devices to customers. - **DMS and asset systems:** supply network and equipment context. Data architecture varies by utility. Some rely mainly on OMS and calls, others combine AMI and SCADA. See [how smart meter data supports outage detection](https://aboutshrey.com/smart-meter-data-grid-operations/) for more. ## SAIDI: System Average Interruption Duration Index SAIDI measures the average total interruption time experienced per customer served over a reporting period. Utilities use it because it reflects how much service time customers lose. Long outages affect it strongly, since each affected customer contributes the full duration. SAIDI = Total customer interruption duration ÷ Total customers served Units are typically minutes per customer (or hours per customer) over the period, depending on reporting convention. A higher SAIDI means more outage time per customer on average. **Limitations:** it is an average, so it can hide local problems, and one major event can dominate it. ## SAIDI Hypothetical Calculation Hypothetical example for illustration. These are not real utility statistics. A utility serves 1,000 customers. Interruptions in the period: - 200 customers × 30 minutes = 6,000 customer-minutes - 300 customers × 20 minutes = 6,000 customer-minutes - 100 customers × 60 minutes = 6,000 customer-minutes Total customer interruption duration = 6,000 + 6,000 + 6,000 = 18,000 customer-minutes. SAIDI = 18,000 ÷ 1,000 = **18 minutes per customer**. ## SAIFI: System Average Interruption Frequency Index SAIFI measures how many sustained interruptions the average customer experiences in the period. It counts interruptions, not their length, so it separates frequency from duration. SAIFI = Total customer interruptions ÷ Total customers served Units are interruptions per customer. Frequency differs from duration because a system can have many short interruptions and a low SAIDI, or few long ones and a high SAIDI. ## SAIFI Hypothetical Calculation Hypothetical example for illustration. A utility serves 1,000 customers: - 200 customers interrupted once = 200 customer interruptions - 300 customers interrupted twice = 600 customer interruptions - 100 customers interrupted once = 100 customer interruptions Total customer interruptions = 200 + 600 + 100 = 900. SAIFI = 900 ÷ 1,000 = **0.9 interruptions per customer**. Here 600 customers were affected, but SAIFI counts every interruption, so repeated outages raise it. The result means that, on average, customers experienced slightly less than one sustained interruption in the period. ## CAIDI: Customer Average Interruption Duration Index CAIDI measures the average duration of an interruption for customers who were interrupted. It is often read as an indicator of how quickly service is restored per interruption. CAIDI = Total customer interruption duration ÷ Total customer interruptions Under the same calculation basis (same period, customers and event set), **CAIDI = SAIDI ÷ SAIFI**. This holds because both indices share the same denominator, total customers served, which cancels out. The relationship applies only when the three indices use identical data and definitions. ## CAIDI Hypothetical Calculation All values are illustrative. Using the SAIDI dataset, where each of the 600 interrupted customers had one interruption: - SAIDI = 18 minutes per customer - SAIFI = 600 ÷ 1,000 = 0.6 interruptions per customer - CAIDI = 18,000 ÷ 600 = 30 minutes. Check: 18 ÷ 0.6 = 30. CAIDI helps distinguish patterns. Consider two illustrative systems with the same SAIDI of 60 minutes: Same SAIDI, different patterns SystemSAIDI (min)SAIFICAIDI (min)Pattern P600.5120Less frequent but longer interruptions Q603.020Frequent but short interruptions ## SAIDI vs SAIFI vs CAIDI Main comparison MetricFull NameMeasuresFormulaTypical UnitMain Question SAIDISystem Average Interruption Duration IndexTotal outage time per customerTotal customer interruption duration ÷ Total customers servedMinutes or hours per customerHow much outage time does the average customer experience? SAIFISystem Average Interruption Frequency IndexInterruptions per customerTotal customer interruptions ÷ Total customers servedInterruptions per customerHow often do customers experience interruptions? CAIDICustomer Average Interruption Duration IndexAverage length of an interruptionTotal customer interruption duration ÷ Total customer interruptionsMinutes or hours per interruptionHow long does an interruption last, on average, for affected customers? - **SAIDI:** how much total outage time does the average customer have? - **SAIFI:** how often do customers experience interruptions? - **CAIDI:** how long does an interruption last, on average, for affected customers? ## MAIFI: Momentary Average Interruption Frequency Index A momentary interruption is a very brief loss of supply, often from a temporary fault cleared by an automatic recloser. Utilities distinguish momentary from sustained interruptions because causes, customer impact and remedies differ. MAIFI = Total customer momentary interruptions ÷ Total customers served Automatic reclosers can restore supply within seconds after a temporary fault, so many events appear to customers as a brief blink. Exact definitions and thresholds vary by reporting framework. A threshold such as five minutes appears in some standards, but other utilities and regulators use different values, so the applicable framework should be checked. ## ASAI: Average Service Availability Index ASAI represents the fraction of time that service is available to customers over a period. It is related to interruption duration but expressed as availability. ASAI = (Customer-hours served − Customer-hours of interruption) ÷ Customer-hours served Conceptually, ASAI is close to 1 minus the share of time customers lose supply. For illustration, if a hypothetical annual SAIDI is 18 minutes, then ASAI ≈ 1 − 18 ÷ 525,600 ≈ 0.99997. It differs from SAIDI by being a ratio near 1, which can make small differences look tiny even when they matter. It offers another perspective on service continuity and is usually read alongside SAIDI. ## Momentary vs Sustained Interruptions Momentary and sustained interruptions compared AspectMomentarySustained DurationVery briefLonger, beyond the defined threshold Typical causeTemporary faults cleared by reclosingPermanent faults, equipment failures Typical responseAutomatic reclosing and protectionCrew dispatch, switching, repair MetricMAIFISAIDI, SAIFI, CAIDI Customer experienceBlink, equipment resetLoss of service A system can have relatively few sustained outages but still experience frequent momentary interruptions, which matters to customers with sensitive equipment. Reporting rules decide which interruptions feed which metric. ## Customer Interruptions vs System Events A system event is not the same as a customer interruption. A single equipment fault may affect one customer, hundreds or thousands, depending on network topology and where the protective device sits. A fault on a lateral may affect a handful of customers, while a fault on a main feeder section can affect many. Reliability metrics therefore count customer interruptions and their durations instead of simply counting equipment failures. This keeps the focus on what customers experienced. ## Planned vs Unplanned Interruptions - **Planned interruptions** are scheduled in advance, such as for maintenance or construction. - **Unplanned interruptions** occur without scheduling, such as faults or emergency outages. Reliability reporting may apply specific exclusions or separate treatment depending on the framework. Some include planned outages, others report them separately. Not every utility treats them identically, so definitions should be confirmed before comparing. ## Major Event Days and Extreme Events Severe events such as major storms, flooding, wildfires where relevant, extreme weather or major infrastructure damage can dominate annual metrics, because they affect many customers for long periods. Some reporting frameworks treat major event days differently, for example by reporting them separately so that underlying day-to-day performance can be seen. No universal exclusion rule applies. Analysts need to know the reporting methodology before comparing metrics across years or utilities. ## Normalized vs Raw Reliability Metrics Reported values may be raw (all events) or adjusted (with certain events excluded or treated separately). Relevant factors include major-event treatment, outlier events and methodology. Two utilities may report different-looking numbers because their methodologies differ, not necessarily because their service differs. When comparing, confirm which events are included, how momentary interruptions are defined and how customers are counted. ## Reliability Metric Data Flow Outage/Event → Detection → Customer Impact → Event Validation → Start Time → Restoration Time → Customer Count → Reliability Database → Metric Calculation → Reporting → Analysis Metrics depend on accurate start and restoration timestamps and correct mapping of customers to the devices that interrupted them. An error of a few minutes across many customers, or a wrong customer count, changes the totals directly. ## Role of OMS An outage management system supports measurement through outage detection, outage location, affected-customer identification, restoration time, event closure and historical outage records. These records are the main input to many reliability calculations. Their quality depends on accurate network models and disciplined closure of events. ## Role of AMI and Smart Meters AMI can improve reliability data through outage detection, last-gasp signals, restoration confirmation, customer-level visibility and event timestamps. Meter-level data can show which customers were actually without power and when supply returned. AMI communication loss does not automatically mean an electrical outage, because a meter may be unreachable while power is on. Utilities need validation and integration with OMS and other systems. See [advanced metering infrastructure](https://aboutshrey.com/advanced-metering-infrastructure-ami/) and [smart meter data quality](https://aboutshrey.com/smart-meter-data-quality/). ## Role of SCADA SCADA provides real-time monitoring, equipment status, alarms, measurements and fault indications, mainly at substations and automated devices. It can identify system events, such as a breaker operation, while AMI and OMS provide customer-level impact information. SCADA, AMI and OMS compared AspectSCADAAMIOMS Primary functionMonitor and control devicesCustomer-level metering and eventsManage outages and restoration Strength for reliability dataSystem event timingCustomer-level power statusEvent records and customer counts LimitationLimited coverage on distributionCommunication gapsDepends on inputs and network model ## Data Quality Challenges - Missing events and duplicate events. - Incorrect timestamps and delayed restoration records. - Incorrect customer mapping and duplicate customer records. - Communication failures. - Network model errors. - Incorrect outage classification. - Manual data-entry errors. Poor data distorts the indices. Missing events understate SAIFI, late restoration records overstate SAIDI, and wrong customer counts shift all three. ## How Utilities Validate Reliability Data - Event validation against multiple sources. - Cross-system comparison of OMS, SCADA and AMI records. - Customer mapping checks against GIS and billing records. - Timestamp validation and restoration confirmation. - AMI validation of power status. - Field verification and OMS reconciliation. - Routine data-quality checks. Automated calculations are only as reliable as the underlying event data. ## How to Interpret SAIDI Simplified interpretation scenarios ScenarioPatternPossible Interpretation AHigh SAIDI, low SAIFIInterruptions may be relatively infrequent but long BLow SAIDI, high SAIFIInterruptions may be relatively frequent but short CHigh SAIDI, high SAIFIBoth frequency and duration are problematic These are simplified interpretations and should be examined alongside network and event context. ## How to Interpret SAIFI A high SAIFI indicates customers experience frequent interruptions, which may relate to fault frequency, protection operation, equipment condition, weather or vegetation. A low SAIFI suggests fewer interruptions, but says nothing about restoration speed. High SAIFI has no single universal root cause, so analysis of causes, locations and devices is needed. ## How to Interpret CAIDI A high CAIDI suggests interruptions take long to restore, which could relate to crew response, fault location, isolation or repair time. A low CAIDI suggests quick restoration per interruption, but it can also fall when many short interruptions are added. CAIDI should be read together with SAIDI and SAIFI. ## Reliability Metric Comparison Example Hypothetical areas with illustrative values. Illustrative area comparison AreaSAIDI (min)SAIFICAIDI (min)Interpretation Area A450.950Moderate frequency and duration Area B1201.0120Interruptions are long Area C603.020Interruptions are frequent but short A lower number is not automatically better without understanding methodology, customer mix, geography, network design, major events and reporting rules. A dense urban network with underground cables and a rural overhead network face different conditions. ## Distribution vs Transmission Reliability Metrics Customer-oriented indices such as SAIDI and SAIFI are primarily used for distribution and customer-service reliability. Transmission reliability uses other concepts, including availability, forced outage rates, contingency performance, system adequacy, security and stability. SAIDI and SAIFI should not be applied to every transmission reliability question, because many transmission events do not interrupt customers directly. ## Reliability Metrics and Grid Resilience Reliability metrics describe service interruption performance. They do not fully capture resilience, which also considers preparation, response, adaptation, recovery, severe events, critical loads and restoration capability. SAIDI and SAIFI are not complete resilience metrics, although they provide useful context. For the distinction, see [grid reliability vs resilience](https://aboutshrey.com/grid-reliability-vs-resilience/). ## Reliability Metrics and Asset Management Reliability Metrics → Problem Identification → Asset / Network Analysis → Maintenance / Capital Planning → Reliability Improvement Metric trends show where to look, for example a feeder whose SAIFI is rising. Metrics alone do not identify root cause. Engineers must analyze outage causes, equipment condition and network design. See [AI predictive maintenance for utility assets](https://aboutshrey.com/ai-predictive-maintenance-utility-assets/) and [what causes power grid failures](https://aboutshrey.com/power-grid-failure-causes/). ## Reliability Metrics and Grid Modernization Utilities may use reliability metrics to evaluate improvements from distribution automation, smart grid projects, AMI, ADMS, FLISR, sensors, predictive maintenance and advanced communications. Technology should be evaluated against actual reliability outcomes and operational objectives, because installation alone does not guarantee better performance. See [grid modernization and real-time visibility](https://aboutshrey.com/grid-modernization-electricity-data-real-time-visibility/) and [what a smart grid is](https://aboutshrey.com/smart-grid-architecture-technologies-utility-applications/). ## AI and Analytics for Reliability Metrics Analytics can support reliability trend analysis, outage clustering, root-cause analysis, anomaly detection, failure-risk analysis, weather correlation, predictive maintenance, reliability forecasting and automated reporting. See [AI for power outage prediction](https://aboutshrey.com/ai-outage-prediction/) and [AI anomaly detection in electric utilities](https://aboutshrey.com/ai-anomaly-detection-electric-utilities/). Analytics can identify patterns, but engineers still need to validate causes and decisions. ## Practical Utility Applications Applications of reliability metrics ApplicationMetric / DataUtility Use Reliability benchmarkingSAIDI, SAIFI, CAIDI with methodologyCompare performance over time or with peers, with care Feeder comparisonFeeder-level indicesIdentify feeders needing investigation District comparisonArea-level indicesAllocate attention across service areas Maintenance prioritizationOutage causes, equipment-related interruptionsTarget inspection and maintenance Capital planningTrends, customer impactSupport investment decisions Outage investigationEvent recordsUnderstand cause and response Restoration performanceCAIDI, restoration timesReview crew and switching effectiveness Regulatory reportingReported indicesMeet reporting requirements Grid modernization evaluationBefore and after metricsAssess project outcomes Reliability improvement programsTargets, trendsTrack progress ## Hypothetical Example: Feeder Comparison Hypothetical feeders with illustrative values. Illustrative feeder metrics FeederSAIDI (min)SAIFICAIDI (min) Feeder A900.75120 Feeder B903.030 Feeder C301.030 Feeders A and B have the same SAIDI, but different patterns. Feeder A has fewer, longer interruptions, which may point to restoration speed, access or fault location. Feeder B has frequent short interruptions, which may point to recurring faults. Feeder C looks comparatively good. A utility might investigate A and B first, using outage causes, equipment condition and field inspection. The metrics show where to look, but they do not identify the cause. ## Hypothetical Example: Before vs After Modernization All values are hypothetical and are not real-world improvement figures. Illustrative before and after comparison PeriodSetupSAIDI (min)SAIFICAIDI (min) BeforeTraditional monitoring, manual restoration1201.580 AfterAMI, automation, improved outage management751.262.5 In this illustration, a lower CAIDI is consistent with faster restoration, and a lower SAIFI could reflect fewer customers affected per event. A real evaluation would also account for weather differences, major events, methodology changes and data improvements. A change in the metrics is not proof of causation. ## Reliability Metric Limitations - Metrics summarize complex events and can hide local issues. - They may be affected by methodology and by extreme events. - Customer weighting matters, since metrics count customers and not load. - Small sample sizes can mislead. - Averages can hide individual customer experiences. - Metrics do not directly identify root cause. - They do not fully capture resilience or every aspect of power quality. ## Common Mistakes - Treating SAIDI as the only reliability metric. - Comparing utilities without methodology. - Confusing SAIFI with SAIDI. - Assuming low SAIFI means good restoration. - Ignoring CAIDI. - Ignoring momentary interruptions. - Ignoring major-event treatment. - Using incomplete outage data. - Treating metric changes as proof of causation. - Comparing different reporting frameworks directly. ## Reliability Metric Improvement Framework 1. **Define the reporting methodology,** including thresholds and exclusions. 2. **Collect complete outage data** from all relevant systems. 3. **Validate event records** for timing and cause. 4. **Map customers correctly** to devices. 5. **Calculate metrics consistently** every period. 6. **Segment by feeder, area or customer class** where useful. 7. **Analyze trends** over time. 8. **Identify root causes** through engineering analysis. 9. **Implement improvements** in maintenance, protection, automation or planning. 10. **Measure results** against the same methodology. 11. **Review methodology periodically** to keep it consistent and appropriate. ## Current, Emerging and Future Maturity of reliability measurement technologies CurrentEmergingFuture / Experimental SAIDI, SAIFI, CAIDIAdvanced analyticsAI-assisted reliability scenario analysis OMSAutomated reliability dashboardsDigital twins SCADAAI-assisted root-cause analysisHighly automated reliability intelligence AMIPredictive reliability analysis Reliability reportingReal-time distribution visibility Adoption varies by utility, and future items are possibilities and not current practice. ## Summary Grid reliability metrics turn customer interruption data into indicators of frequency, duration and availability. SAIDI measures outage time per customer, SAIFI measures interruptions per customer, CAIDI measures average interruption duration, MAIFI covers momentary interruptions and ASAI expresses availability. Accurate data collection, validation and a clear methodology are essential, and each metric has limits. Used together with network context, they support operations, maintenance, planning and regulatory reporting. ## Conclusion SAIDI, SAIFI, CAIDI and related metrics give utilities different views of interruption performance. The most useful reliability analysis combines these metrics with outage data, network context, asset information, operational records and engineering judgment. No single metric defines grid reliability, and the numbers are most valuable when everyone understands how they were produced. ## Frequently Asked Questions About Grid Reliability Metrics ### What are grid reliability metrics? Grid reliability metrics are standardized indices that quantify how often and how long customers experience interruptions, and how available service is. Common examples are SAIDI, SAIFI, CAIDI, MAIFI and ASAI. They convert outage records into numbers that can be tracked, reported and used in planning. No single metric describes reliability completely. ### How do utilities calculate grid reliability metrics? Utilities record each interruption’s duration and the customers affected, then apply formulas. SAIDI is total customer interruption duration divided by customers served. SAIFI is total customer interruptions divided by customers served. CAIDI is total duration divided by total interruptions. Definitions, thresholds and exclusions follow the applicable reporting framework, so methodology should be confirmed. ### Why are grid reliability metrics important? They give utilities, regulators and planners a consistent way to track service performance, find weak areas and compare results over time. They support maintenance prioritization, investment decisions and evaluation of modernization projects. Their value depends on data quality and understanding methodology, since averages can hide local problems. ### What are the main grid reliability metrics? The main metrics are SAIDI (average interruption duration per customer), SAIFI (average interruptions per customer) and CAIDI (average duration of an interruption). MAIFI covers momentary interruptions, and ASAI expresses service availability. Together they describe frequency, duration and availability, and are usually read as a set. ### What are the challenges of measuring grid reliability? Challenges include missing or duplicate outage events, wrong timestamps, incorrect customer mapping, network model errors, communication failures and manual entry mistakes. Differences in methodology, such as major-event treatment and momentary thresholds, complicate comparison. Utilities address them through validation, cross-system reconciliation and consistent reporting rules. ### Where are grid reliability metrics used? They are used in regulatory reporting, benchmarking, feeder and area comparison, maintenance prioritization, capital planning, outage investigation and restoration review. Utilities also use them to evaluate grid modernization projects. They are primarily customer-focused distribution measures, while transmission reliability uses other concepts such as availability and contingency performance. ### What is the difference between SAIDI, SAIFI and CAIDI? SAIDI measures average total outage time per customer, SAIFI measures average number of interruptions per customer, and CAIDI measures average duration of each interruption. Under the same basis, CAIDI equals SAIDI divided by SAIFI. Together they show whether performance issues come from frequent interruptions, long interruptions or both. ### How do utilities use reliability metrics to improve grid performance? Utilities track metrics by feeder and area, identify where performance is weak, and investigate causes with outage records, asset condition and field inspection. Findings guide maintenance, protection, automation and capital projects. Results are measured against the same methodology. Metrics show where to look, but engineers determine causes and solutions.