Leakage remains one of the UK water sector’s most persistent challenges. The pressure to reduce it is growing, but the real question has shifted: not just how much, but how. Every District Metered Area (DMA) behaves differently. Two DMAs that look identical on paper can respond in completely different ways to the same intervention, because leakage is shaped by a tangle of factors, like pressure regimes, pipe condition, demand patterns, local operational practices, that interact in ways that are difficult to unpick.
Skilled practitioners have always navigated this complexity through experience and judgement. But with a widening skills gap and mounting targets, experience alone cannot scale fast enough. The industry needs structured, data-informed approaches that capture what good practitioners already know and make it accessible, repeatable and transparent.
This is exactly the kind of problem Artesia was set up to solve. We sit at the intersection of rigorous research and practical water industry operations, and our strength lies in taking complex academic methodologies and turning them into tools that practitioners can actually use.
Bringing Bayesian Networks to leakage management
The recent UKWIR project Understanding Factors Contributing and Affecting DMA Leakage, delivered by Artesia and Dayworth Consulting, is a good example. The project drew on research by the University of the West of England (UWE) into Bayesian Networks: statistical models that map how multiple variables influence one another, capturing both direct and indirect relationships. UWE’s work showed how these networks could untangle complex, interrelated systems. We recognised immediately that the same approach could help explain why leakage behaves so differently from one DMA to the next.
But recognising a method’s potential and making it work in practice are two very different things. Translating Bayesian Networks from an academic context into something a leakage manager can use on a Tuesday morning required deep understanding of both the mathematics and the operational reality. It meant extensive engagement with practitioners across the sector, careful interrogation of imperfect datasets, and the integration of expert judgement alongside quantitative analysis. The model had to work with incomplete data, because in practice, asset and operational records are always fragmented and it had to reflect the messy, context-specific nature of real leakage management.
From model to practice
The outputs were practical playbooks: structured guides that help teams understand DMA-specific leakage behaviour, compare intervention options, and run “what-if” scenarios before committing resources. They support clearer, more consistent decision-making without replacing the practitioner’s role; instead, they give experienced operators better tools to work with, and help less experienced teams build understanding faster.
This is the kind of work Artesia does best. The water industry is not short of good research, and it is not short of operational expertise. What it often lacks is the bridge between the two – people who can read an academic paper, understand the method, see its practical application, and then do the hard work of adapting it for use in the field. That translation process demands a rare combination of analytical depth and industry knowledge, and it is where we add most value.
The Bayesian Network approach is designed to evolve. As new data becomes available and new insights emerge, the models can be updated and refined. This is not a one-off study. It is a framework that grows with the industry’s understanding, helping the sector move towards smarter, more targeted leakage management over time.
Here is the UKWIR report on Understanding factors contributing and affecting DMA leakage.
