Water companies in England and Wales are operating with a fundamental blind spot. A significant portion of their customer base - unmetered households billed on rateable value - consume water with no direct measurement in place. These customers pay a fixed charge regardless of Ahow much they use. The company receives no consumption signal. Demand management, efficiency targeting, and metering strategy are all working, to varying degrees, in the dark.

This matters more now than it did a decade ago. Per capita consumption targets, water resource management plans, and Ofwat's increasing focus on demand reduction have raised the stakes on understanding where water actually goes. Companies that cannot identify their highest-using unmetered customers cannot engage them efficiently, cannot make evidence-based decisions about meter installation priority, and cannot design water efficiency campaigns that reach the households where they will have the greatest impact.

AI-driven customer profiling is changing that. By building predictive models trained on the rich consumption data now available from metered customers, water companies can infer probable usage levels across their unmetered estate - identifying the properties most likely to be high consumers, and targeting intervention accordingly.

Why Unmetered Demand Is a Strategic Problem

The proportion of unmetered households varies significantly across water companies, reflecting differences in metering programme history, geography, and regulatory approach. For companies still carrying a substantial unmetered base, the challenge is not just operational - it is strategic. Water resource management plans are built on demand forecasts, and demand forecasts for unmetered properties rely on assumptions rather than measurement. The further those assumptions diverge from reality, the less reliable the plan.

The commercial dimension is equally significant. Unmetered customers who use substantially more water than the tariff assumes are effectively receiving a subsidy - their bill does not reflect their consumption. At scale, this represents meaningful unrecovered revenue and distorts the economic signal that metering is designed to create. Where universal metering is the long-term objective, understanding the unmetered estate is a prerequisite for planning the transition intelligently.

Regulatory context:

Ofwat's PR24 determinations place demand reduction at the centre of long-term water resource planning. Companies are expected to demonstrate credible strategies for reducing per capita consumption toward 110 litres per person per day by 2050. Identifying and engaging high users in the unmetered base is a direct lever for delivering against that target.

How AI Profiling of Unmetered Customers Works

The modelling approach draws on a well-established principle: metered customers in similar properties, with similar household characteristics, in similar locations, consume water in broadly predictable patterns. By analysing the relationship between consumption and a wide range of property and demographic variables in the metered population, AI models can generate probabilistic consumption estimates for unmetered properties where those same variables are observable.

Building the model from metered data

Smart meter and AMR data from the metered estate provides the training foundation. The model learns which combinations of property type, size, occupancy, garden presence, location, and other variables are associated with high, medium, and low consumption levels. The more metered data available - and the more granular - the stronger the predictive signal.

Enriching with third-party property and demographic data

Consumption alone is not sufficient. Property characteristics - floor area, number of bedrooms, garden size, age of build - are strong predictors of water use and are widely available through land registry, council tax, and commercial property data sources. Demographic indicators at postcode or LSOA level add further signal: household size estimates, age profile, deprivation indices, and garden ownership rates all correlate with consumption patterns and are incorporable without requiring individual-level data that raises privacy concerns.

Generating the unmetered probability distribution

Trained on the metered population, the model is applied to the unmetered estate to generate a probability distribution of likely consumption for each property. The output is not a single consumption figure - it is a ranked risk profile, identifying the properties most probably consuming well above average, with a confidence level attached to each estimate. This probabilistic framing is important: the model is not claiming to know what an unmetered property uses; it is identifying where the evidence most strongly points to high consumption.

VE3 perspective:

The predictive accuracy of unmetered profiling models improves substantially as smart meter coverage in the metered base increases. Companies with active smart meter rollout programmes should treat that data asset as directly valuable for unmetered modelling - not just for metered customer management. Designing the data architecture to support both use cases from the outset reduces the cost of building the unmetered capability later.

What the Profiling Enables

Targeted metering strategy

Rather than installing meters based on geography or rolling programme convenience, companies can use propensity scores to prioritise meter installation at properties most likely to be high consumers. This concentrates metering investment where it will generate the greatest demand reduction, the strongest revenue recovery, and the most valuable data - accelerating the return on metering programme spend.

Water efficiency campaign targeting

Generic water efficiency campaigns reach everyone and move few. Targeted campaigns - designed for specific customer segments, delivered through appropriate channels, and focused on the behaviours most likely to drive consumption in high-use properties - consistently outperform. AI profiling makes that targeting possible in the unmetered base, directing efficiency advice, device provision, and behavioural engagement to the households where they will have the highest impact on consumption.

Demand forecasting and resource planning

Profiled consumption estimates for the unmetered estate improve the quality of demand forecasting inputs used in water resource management plans. Rather than applying uniform per-property consumption assumptions, companies can build more nuanced demand models that reflect the probable distribution of consumption across the unmetered base - reducing forecast uncertainty and supporting more robust long-term planning.

Tariff and programme equity analysis

Understanding the probable consumption distribution in the unmetered base also allows companies to assess the equity implications of rateable value billing - identifying the degree to which low-consuming households are cross-subsidising high consumers, and informing the policy case for accelerated metering or tariff reform.

Data Privacy and Ethical Considerations

Any programme that generates inferred data about individual customers requires careful design from a privacy and ethics standpoint. The General Data Protection Regulation (GDPR) and the UK Data Protection Act 2018 apply fully to AI profiling of customer data, and water companies must be able to demonstrate a lawful basis for processing, proportionality of use, and appropriate data minimisation.

In practice, well-designed unmetered profiling programmes handle this through two principles. First, the model operates at the property level rather than the individual level - the output is a property consumption probability, not a personal profile. Second, the enrichment data used is aggregated or anonymised at postcode or area level, rather than sourced from individual-level data sets. These design choices preserve analytical utility while materially reducing privacy risk.

Transparency with customers is also good practice. Where profiling outputs are used to inform direct customer engagement - an invitation to participate in a metering trial or an efficiency scheme, for example - the basis for that targeting should be explainable and fair. Customers identified as probable high users should not face adverse consequences from that classification; the engagement should be supportive and voluntary.

How VE3 Delivers Unmetered Customer Profiling

VE3 Global brings together data science, customer analytics, and water sector experience to build unmetered profiling capabilities that are accurate, privacy-compliant, and operationally integrated. Our approach covers:

1. Metered consumption modelling: building predictive models from smart meter and AMR data, enriched with property and area-level demographic data, validated against holdout samples before deployment.

2. Unmetered estate scoring: applying models to the full unmetered customer base to generate ranked consumption probability profiles, with confidence intervals and segment classifications.

3. Targeting workflow design: translating model outputs into practical targeting logic for metering programmes, efficiency campaigns, and demand management initiatives - with reporting that tracks conversion and impact over time.

4. Privacy and governance framework: designing the data processing architecture and documentation required to demonstrate GDPR compliance and proportionality of use.

Conclusion: Inference Is Not a Compromise - It Is a Starting Point

Not knowing what unmetered customers consume is not an acceptable long-term position for a water company with serious demand reduction ambitions. AI profiling does not replace metering - it makes the path to metering smarter, and makes everything that comes before metering more effective. The data to build these models is already sitting in the metered estate. Applying it to the unmetered base is one of the clearest examples in the sector of using existing data assets to generate new operational and strategic value.

About VE3 Global

VE3 Global is a UK-based technology and enterprise AI consultancy, partnering with water companies, regulated utilities, and public sector organisations to deliver AI, data, and digital transformation programmes that create measurable operational and commercial value. With offices in London and Pune, VE3 combines deep sector knowledge with cutting-edge AI capability to help clients navigate the full journey from data strategy to production AI deployment.