Over the past three decades, Latin America and the Caribbean has recorded electricity losses averaging approximately 17% of generated energy. For the 15 countries where separate transmission and distribution data were available, distribution accounted for around 80% of recorded losses, with an estimated annual financial cost to distribution companies of US$9.6–16.6 billion. [1]
Performance varies substantially between utilities. In the 2019 distributor data analysed by the IDB, some utilities reported total losses below 4%, while others reached approximately 47%. [2]
This represents a substantial volume of energy that has been produced and delivered but never fully converted into revenue.
In 2025, Brazil’s distribution losses represented 14.3% of injected energy. Technical losses accounted for 7.2%, or 45.2 TWh, while non-technical losses represented another 7.1%, or approximately 45 TWh. [3]
The problem is also highly concentrated. In 2024, ten distributors accounted for 74% of Brazil’s non-technical losses, while Light and Amazonas Energia alone represented 34.1%. ANEEL estimated the annual cost of actual non-technical losses at approximately R$10.3 billion. [4]
These figures indicate that utilities need more than broad fraud-detection programmes applied equally across the entire customer base. They need to understand where losses occur, what causes them and which cases should be prioritised for further investigation.
Electricity theft is the most visible form of NTL, but it is only one part of the problem. For operational purposes, non-technical losses can be grouped into four broad categories:
Unauthorized consumption and meter tampering
Illegal connections, meter bypassing, reverse meter operation, physical manipulation, unauthorized network extensions and reconnection after disconnection.
Unmetered and incorrectly configured consumption
Unregistered consumption points, malfunctioning meters, incorrect meter or transformer configuration and equipment operating outside its intended measurement range.
Incorrectly recorded consumption and process errors
Inaccurate readings, registration errors, incorrect meter records and delayed updates related to customers, connections or disconnections.
System, tariff and billing inconsistencies
Incorrect tariff classifications, delayed master-data exchange, synchronization issues and consumption that is not correctly transferred, allocated or billed. [5]
Treating every anomaly as theft can produce false positives, unnecessary inspections and avoidable customer disputes. Different causes require different investigation methods and corrective actions.
Non-technical loss is often approached primarily as a detection problem: identify more anomalies, generate more alerts and inspect more customers.
In practice, the greater challenge is converting detected anomalies into cases that investigators can understand, prioritise and resolve.
Utilities may already possess much of the evidence needed to identify an irregularity. However, that evidence is often distributed across metering, billing, customer, network and operational systems. Investigators may need to retrieve and compare information manually across several disconnected applications.
This increases investigation time, creates inconsistent assessments and delays decisions on corrective actions or field inspections.
Detection alone is therefore insufficient. Without structured filtering, risk scoring and prioritisation, utilities can be left with long lists of potential cases of uneven quality. Field teams may then be dispatched to cases with a low probability of confirmation or limited recovery potential.
The problem is not necessarily a lack of alarms, but inability to convert fragmented information into a smaller and prioritized pipeline of investigations.
Addressing this challenge does not necessarily require replacing the utility’s existing systems. It requires a connected process that uses available data and links detection, triage and investigation through to a documented outcome.
Our Comping NTL solution framework is designed to support this process by integrating and preparing information from relevant utility systems and external sources.
It combines configurable business rules with statistical and analytics models to detect suspicious cases, generate alerts and assign risk scores. The framework also supports structured triage, investigation workflows and dashboards, case management, reporting and auditability. [5]
Comping NTL addresses the specialized process of detecting and managing non-technical losses. For utilities seeking to build on that foundation, Thaora expands the approach across a broader set of metering, consumption and distribution-intelligence use cases. Thaora, a software platform for consumption and distribution intelligence. can serve as an integration and aggregation layer for smart-metering data, consolidating information from available sources and providing standardized inputs for Comping NTL and additional analytical use cases. This creates a more reusable data foundation and reduces the need to establish separate integrations for each new initiative.
The same platform can support broader use cases such as consumption analysis, operational performance monitoring, and improved grid visibility down to the secondary-substation level.
Rather than treating NTL as an isolated analytics project, utilities can use Thaora to extend the same data-driven approach across a wider set of consumption, network and operational challenges. [5]
The framework can identify patterns such as unusual consumption changes, missing or inconsistent readings, consumption after disconnection, threshold exceedances and anomalies across related customers, meters, locations or network segments. [5]
Cases can then be reviewed and prioritised using criteria such as risk score, detection scenario, consumption value and defined investigation rules.
The objective is not to produce the longest possible list of suspicious customers, but to identify the next best investigation.
Business value is generated by connecting three phases:
Investigation results can then be returned to the analytical process. Confirmed irregularities, false positives and unresolved cases provide inputs for periodic model review and tuning, helping improve detection accuracy over time. [5]
In one utility implementation, this connected process yielded:
These outcomes are implementation-specific and depend on the utility’s available data, baseline detection performance, investigation processes and operating model.
Reducing non-technical losses is not just a fraud-detection initiative, it is also a revenue-protection programme, a data-quality programme and a distribution-modernisation programme. It affects utility profitability, investment capacity, customer trust and the ability to provide reliable service.
Most utilities do not need more disconnected alarms. They need a structured view of where losses occur, which cases deserve attention and which action is most likely to produce value.
NTL reduction is not primarily about detecting more anomalies.
It is about turning available data into fewer, better and more actionable cases.
The next step in NTL management is not mass detection. It is intelligent case conversion.
References
[1] Yépez-García, Ariel, and Raúl Jiménez Mori, eds. The Economics of Electricity Losses in Latin America and the Caribbean. Inter-American Development Bank. 2024.
[2] Bonzi Teixeira, Augusto, Eric Fernando Boeck Daza, Michelle Carvalho Metanias Hallack, Mariana Weiss, Yuri Daltro, Arturo Daniel Alarcón Rodríguez, and Leopoldo Montañez. Electrokit: Power Utility Toolkit—Electricity Loss Reduction. Inter-American Development Bank. 2021.
[3] Agência Nacional de Energia Elétrica. ANEEL Divulga Relatório sobre Perdas de Energia Elétrica na Distribuição: Dados de 2025. 2026.
[4] Agência Nacional de Energia Elétrica. Relatório Traz Níveis de Perdas Técnicas e Não Técnicas em 2024 no Sistema de Distribuição de Energia. 2025.
[5] Internal proprietary insights derived from real-world experience with existing clients
