Percentages don’t work for

Tracking NRW and leakage reduction

by Allan Lambert, Water Loss Research & Analysis Ltd

International water loss professionals who are now fully aware of the many pitfalls of using %s don’t  recommend %s of System Input Volume to compare performance in management of Non Revenue Water and leakage between Water Utilities (metric benchmarking).  However, some traditional users of %s still assume that %s can be valid KPIs for setting targets and tracking reduction of NRW and leakage within an individual Utility (process benchmarking). The following examples show just how misleading this assumption can be.

Two Examples from Germany

For more than 20 years, DVGW (the largest technical standards group for Water and Gas in Germany) has recommended against the use of % as a performance measure for NRW and Water Losses. DVGW was the first National organisation to adopt the IWA Water Balance, for Germany, after it was first published in late 1999; they continued to use their traditional measure (Specific Loss – m3/km mains/hr) as a KPI for Water Losses. However, the latest draft of the new DVGW Guidelines W392, expected to be published in 2014, commits to using ILI as the best performance indicator for Real Losses, with the traditional Specific Loss figures also to be quoted.

Apparent losses in Germany are very low due to high customer metering standards and direct pressure systems without customer storage tanks. Customer meters are generally located within the entry to buildings and Utilities have powers to implement rapid repairs on service pipes between the property line (equivalent to curb stop in USA) and the customer meter – the average length of this pipe is around 7 metres (23 feet).

Water Losses have generally been low in former Western Germany due to high technical standards for distribution system infrastructure. An initial (as yet unpublished) recent study by Lambert and Koelbl, based on analysis of consultants reports of aggregated water balances for large percentages of Utilities in 4 of the 16 German Lander (States) produced figures for Real Losses of:

57, 66, 68 and 85 litres/connection/day and ILIs of 0.74, 0.93, 1.08 and 1.10

The lowest of these ILIs is in a geographical location where most leaks surface naturally, so an ILI less than 1.0 is feasible. The pooled ILIs for the other three States are effectively at ILI = 1.0, allowing for random errors in the water balance calculation. Clearly, water loss management in much of Germany is of a very high standard. However,  such Utilities with ILIs close to 1.0 – the IWA ‘Unavoidable’ or ‘Technical Minimum’ level of Real Losses, allowing for key local factors of connection density, meter location and average system pressure  – have negligible headroom to further reduce Real Losses if pressure is already being effectively managed.

Customer-side demand management initiatives in Germany over the past 20 years – more effective washing machines, dishwashers, low flush toilets and shower heads – and decreasing population in the former Eastern Germany –  has reduced consumption and system input volume significantly, and the data for WasserBerlin (see graph below) shows that if water losses had been held at a constant volume of 10 Million m3/year since 1990, % losses would have almost doubled from 3% to 5.1% of System Input Volume as a result of this progressive reduction in consumption.

Source: Realistic Balance of Water Losses and the Practical Use of Water Loss Indexes: Article by Lars Tennhardt, Berlin WaterWorks, in Energie: Wasser-Praxis, 10/2012. English translation by Koelbl J and Schwaller J.

Source: Realistic Balance of Water Losses and the Practical Use of Water Loss Indexes: Article by Lars Tennhardt, Berlin WaterWorks, in Energie: Wasser-Praxis, 10/2012. English translation by Koelbl J and Schwaller J.

A  further German example of the futility of tracking progress using % of system input volume can be found in an article on water loss management in Albstadt, a town in South West Germany with permanent leakage network monitoring; this example is quoted in an article in ‘Water Loss Detectives’, May 2012 Issue, a magazine published by AQUATIM, Romania ( To quote from the article, ‘Current NRW is 20% (500,000 cubic metres), which has increased from 10% five years ago. NRW percentage is due to a 50% reduction in total consumption, our water loss has remained constant.’

The statistics from more than 6000 Water Utilities in Germany are published in summarised form by DVGW (Statistiches Bundesamt), the last set being for 2007 data, published in 2009. No performance indicators are shown in that Report. However, despite the long-standing DVGW recommendation not to use %s as a Performance Indicator for Water Losses, some individuals and organisations use this data to calculate %s.

A National Report by VEWA Consultants in 2010 for BDEW (the German Association of Energy and Water Industries) quotes 6.5% water loss for Germany, based on the 2007 Statistisches Bundesamt data. In contrast, an enquiry to DVGW by Koelbl produced a reply that the 2007 data, if calculated as a %, would represent 8.9% water losses. Further enquiries to BDEW by Koelbl to seek explanation for the difference met with no response – twice. It is strongly suspected by Koelbl and Lambert that the lower BDEW figure includes double counting of the volumes of imported water. But in fact, as there are several different valid ways of doing the % calculation, even without assumption errors, several different %s can be quoted for the same set of data, by including or excluding raw water abstracted, water imported, etc., and so figures can easily be manipulated simply by adjusting the method of calculation, if nobody is prepared to challenge the validity and method of the calculations.


An Example from Australia

During the first 10 years of this century, much of Australia experienced a sustained and intense drought (the millennium drought) which created critical water supply conditions in most of the country. Yarra Valley Water, the largest Distribution Utility in the Melbourne Area, employed a wide range of modern water loss management techniques to drive down their Non-Revenue Water over this period.

Water Leakage and Pressure Management Australia

The Water Services Association of Australia ceased using NRW %s by volume in 2003, when they adopted the ILI. However, YVW’s regulator, the Victoria Essential Services Commission, still requires NRW to be quoted as a % of Water Supplied. In 2013, a separate article described YVW’s 2010/11 performance as ‘a long journey to 12% NRW’. For anyone who considers that NRW% is a reliable indicator for tracking progress, in NRW reduction, it is instructive to delve a little deeper into the YVW data.

Billed Water Consumption and Real Losses

The first graph above shows that NRW % and Real Losses % hardly changed at all between 2000/01 and 2010/13! NRW% fluctuated between 9.7% and 14.1%, averaging 12.5%; Real Losses % varied between 7.3% and 11.5%, averaging 9.5%. Those who consider that %s of Water Supplied can be used to track progress in management of NRW and Real Losses would, not unreasonably, conclude that YVW were not making significant progress. The second graph above shows how wrong they would have been. During the worst of the drought from 2003 to 2012:

  • the YVW system was expanding and the number of service connections increased by 11% (from 505,000 to 553,000)
  • demand management measures reduced billed consumption  by 31% (from 451 to 312 Ml/day)
  • Non-Revenue Water volume/day was reduced by 41% (from 64 to 48 Ml/day)
  • Real Losses volume/day were reduced by 44% (from 50 to 28 Ml/day).

The activity of tracking progress in performance within a single Utility is known as Process Benchmarking, or Performance Improvement. As the majority of NRW and Real Losses in well managed urban systems are associated with service connections, the best simple performance indicator for tracking progress in litres/service connection/day. For comparing Real Losses management performance between Utilities (Metric Benchmarking, or Performance Assessment), the Infrastructure Leakage Index (ILI) is recommended, as it takes account of differences in density of connections, customer meter location on the service connection, and average pressure. The next two graphs show the tracking of these two KPIs.

NRW Real Losses - Water Leakage and Pressure Management

The Process Benchmarking KPIs of litres/service connection/day clearly show that both NRW and Real Losses were progressively driven down by YVW, year by year, over the worst period of the drought between 2003 and 2011:

  • NRW was reduced by 46%, from 127 to 68 litres/connection/day
  • Real Losses were reduced by 48%, from 99 to 51 litres/connection/day.

The Infrastructure Leakage Index also showed progressive reduction of 46%, from 1.3 to 0.7.  Using this KPI for its intended purpose of comparing Utilities’ performance, it can be seen that by 2011-12 (the last published National Data set), many Australian Utilities were achieving an ILI of 1.0, which represents the international ‘gold standard’ of technical leakage management. Yet YVW managed to achieve an ILI of 0.7, an outstanding performance and the best of any of the largest Australian Utilities.

Australian Water Utilities Water Leakage and Pressure Management


So what can we conclude from the BerlinWasser, Albstadt and Yarra Valley Water experiences? The two German Utilities have already publicised a real problem with use of %s for comparing and tracking performance. This problem no doubt occurs in other German Utilities, and DVGW W392 standards continue to recommend strongly against the use of %s, with good reason. However, BDEW ignores that good advice and will increasingly be faced with the problem of explaining why % water losses in Germany keep rising. BDEW’s lack of any response to valid questioning of their calculations does not seem to encourage an open recognition that there is a problem.

Water Services Association of Australia ceased to use %s in 2003, and adopted the ILI and litres/connection/day in 2003. However, the National Water Commission Annual Reports continue to use litres/connection/day for comparisons between Utilities (metric benchmarking), when if they used the ILI for that purpose they would find that the Australian Urban Water Industry was a World Leader in NRW and Real Losses management.  But litres/connection/day, as we have seen from the YVW example, is the ideal KPI for tracking progress within most individual Utilities.

Any State or National organisation that chooses to use %s to track performance in management of Utilities with low water losses will, in the 21st Century, almost inevitably face these problems. But as we have seen from these German and Australian examples, ‘%s just don’t work’. Those who misguidedly continue to use them, by implying to lay persons that %s have some technical merit, are actively sabotaging the efforts of water professionals whose purpose is to save water rather than to perpetuate misconceptions.

Acknowledgements: Joerg Koelbl, Lars Tennhardt, Julie Schwaller, Yarra Valley Water

Leakage Tracking – Percentages