clm-2e3e132a729b4e1e
quantitative
lost 72% of what it produced
Phnom Penh cut water losses from 72% to 6% with meters, a customer census and enforced bills. In 2024 its own figure was 27.27%. What changed is unclear.
This overview was generated with AI from the public sources listed below and checked against them before publication. It has not had a complete human fact-check — treat it as a starting point and follow the sources.
31 sourced claims on this page, none individually verified yet — turn on “Sources in the text” to see each one.
How this atlas is made · What “last verified” means · What “source-checked” means · The full record behind this page
✎ Corrected 25 Sep 2026 Evidence grade changed: independent → peer-reviewed. · 3 revisions in all What changed →
Part of 1001 Smart Cities — 1,216 documented projects · 🏙️ 1 more in Phnom Penh · 💧 Water guide · 📡 Data & IoT projects
Phnom Penh’s water system came out of the Khmer Rouge years and the decade after them in a state that is hard to overstate. The city had pipes, some of them seventy years old, and a utility on paper. It did not have a reliable list of its own customers, a way to measure what they used, or staff with a reason to care. Non-revenue water topped 72% and revenues collected barely covered half of operational costs, and those who were connected got water for only part of the day.
The loss figure is often read as leakage, and some of it was. But the first problem was that the utility did not know who its customers were. In 1993 only 3,391 of its 26,881 connections had a meter, so most bills were estimates with no link to what a household used. When a hundred staff walked the city house by house in 1994, they found 13,901 households receiving water but no bill, and 12,980 being billed for water they never received.
The other half of the loss was organised. According to the utility’s own later account, an illegal connection cost about $1000 and could be arranged by middlemen linked to the utility's top managers, while residents drilled into the mains to fill shared underground tanks. A connection through the front door was close to impossible; one through the back door was a business.
The repair began with the ledger, not the control room. Metering of every connection began in 1994, and by 2001 all connections were metered, with older Class B meters swapped for more accurate Class C ones from 1997. The billing database was computerised with a French grant by 1996. None of this is sophisticated technology. It is the precondition for every sophisticated thing that came after, because a network model is only as good as the consumption figures fed into it.
Telemetry followed sooner than the usual telling suggests. By 2008 the distribution network was divided into 41 zones, each equipped with a pressure and flow data transmitter sending online data to find large leaks, and by 2010 process control had been shifted to a SCADA system. What came much later was integration: a GIS platform combining SCADA with the Sense Water loss-management system came after 2015, and spot-billing from mid-2017. The order matters for any city trying to copy this. The zones and the sensors made leaks visible; the meters and the census made the numbers mean something.
Most accounts of this turnaround spend more time on people than on pipes, and for good reason. Staff had been paid little and promoted by connection. The new management raised pay sharply, tied it to performance, and removed staff found helping illegal connections. Payment was made universal, which was the politically harder part: the army initially refused to have any meter installed, and in 1997 the Prime Minister publicly declared that every person and institution had to pay their water bills. A block tariff introduced that year and a second increase in 2001 were set, according to the evaluation, after customers had already seen the service improve.
The record on the poor is more mixed than the headline, and the peer-reviewed evaluation says so itself. A first scheme sold bulk water to community representatives for resale, and the representatives mostly sold water to the poor at 10 to 20 times the subsidised rate. A World Bank-funded connection programme managed only 101 poor households in 1999. The model that eventually worked subsidised the connection itself: poor households could receive a subsidy of 30, 50, 70 or 100% of the connection fee, and connections to poor households rose from 101 in 1999 to 17,657 by the time of the evaluation. The same four tiers are still in use today.
| When | What happened | Who says so |
|---|---|---|
| 1993 | Loss rate 72%; Ek Sonn Chan appointed general director | Operator; award citation |
| 1994 | House-to-house customer census; metering of all connections begins | Operator; peer-reviewed |
| 1996 | Decree No. 52 grants administrative and financial autonomy; first leak-management group | Peer-reviewed; ADB; operator |
| 1997 | Block tariff introduced; Prime Minister orders everyone to pay | Peer-reviewed |
| 2001 | All connections metered; second tariff increase | Peer-reviewed |
| 2008 | Loss rate 6.19%; network in 41 telemetered zones | Peer-reviewed (from PPWSA records); ADB |
| 2010 | Stockholm Industry Water Award; utility cites 5.9% and aims for 4% by 2020 | Award announcement |
| Apr 2012 | First company listed on the Cambodia Securities Exchange | Operator; press |
| After 2015 | Integrated GIS platform with SCADA; spot-billing from mid-2017 | Operator |
| 2023 | Loss rate reported at 8.75% | Operator |
| 2024 | 9.75% in the first quarter, 27.17% in the second, 27.27% for the year | Operator |
| 2025 | 25.55%; board asks for a fuller loss report, internal audit reviews the calculation | Operator |
The low point did not hold. When it won the Stockholm award in 2010, the utility set out its next ambition in public: reducing water loss from the current 5.9 percent to 4 percent by 2020. By 2023 the operator’s own annual figure was more than double that target, though still in single digits: in 2023 the water loss rate was kept in single digits at 8.75%. Over the same period the network grew from 288 kilometres in 1993 to several thousand, and the service area spread into Takmao and beyond.
Then the series breaks. PPWSA’s quarterly filings to the stock exchange report 9.75% in the first quarter of 2024 and 27.17% in the second, and the annual reports give 27.27% for Phnom Penh and Takmao in 2024 and 25.55% in 2025. The only cause they name is mundane: road repair and widening across Phnom Penh is one of the many reasons for leaking pipes. The 2025 report is unusually frank about the state of the measurement itself. The board noted that the latest quarterly loss report was not yet a comprehensive one, and internal audit was assessing the methods used to calculate the rate. The new target is 15% by 2030 in the chairman's outlook, and 15% by 2027 in the board-approved action plan.
The utility is also a much larger machine than the one that won the award. The Bakheng plant produced 102,751,262 cubic metres in 2025, with a third phase scheduled for completion in the fourth quarter of 2027. More water through a longer, older network is exactly the condition under which losses grow, which is why the comparison with Johannesburg or Istanbul now runs in both directions.
Three things, and they are the ones that decide how this case should be read today.
First, whether the 2024 jump is water or accounting. A jump from under 10% to over 27% in a single quarter is faster than pipes age. It could reflect a change in how non-revenue water is measured, a change in the area it covers, a real deterioration, or some mix. The reports do not say, and the board’s own request for a more complete loss report suggests the utility is still working that out.
Second, how independent the celebrated low really is. The evaluation's charts are marked as drawn from PPWSA records. That does not make them wrong, and the authors examined the utility closely. But the 72%-to-6% arc has never, as far as this atlas can find, been reconstructed from an external water balance. The Stockholm Industry Water Award is a prize with a jury citation, not an audit.
Third, what the poor pay today. The connection subsidies are documented; household bills as a share of income in the peri-urban areas the network now reaches are not, in any source this entry cites.
Non-revenue water fell from 72% in 1993 to 6% in 2008, with 24-hour supply reaching about 90% of a city of 1.3 million and no state operating subsidy. It did not stay there: PPWSA's own reports give 8.75% for 2023, 27.27% for 2024 and 25.55% for 2025, and internal audit is reviewing how the rate is calculated.
Two clocks, kept apart: when this entry’s sources were published, and what has happened to its record in the atlas. Documented span: since 1993. 14 of 16 sources carry a publication date, from 30 September 2008 to 7 April 2026.
The 1993–2008 turnaround rests on one peer-reviewed paper, Biswas and Tortajada in the International Journal of Water Resources Development (2010), read in the authors’ open copy because the publisher’s page blocks automated readers; on a 2008 Asian Development Bank case study built largely on interviews with the utility; on a 2010 IUCN study co-authored with PPWSA’s own leadership; and on the 2010 Stockholm Industry Water Award announcement. Two sources that look external are written by the operator: the 2021 article in IWA’s magazine The Source is by PPWSA’s then director general, and the 2025 IWA case story by PPWSA staff. Everything after 2010 comes from PPWSA’s annual and quarterly reports and its website, and the website’s loss page still stops at 2023. Small disagreements between sources are left visible rather than resolved: the 1994 census found 13,901 unbilled households in the paper and 13,722 on the utility’s website; 2023 appears as 8.75% in the annual report and 8.93% in the IWA piece; and the operator dates its third tariff increase to May 2017 in one account and to 1 January 2020 in its annual report. Open questions: what drove the 2024 break in the loss series; whether any external water audit of the utility has been published; and what water costs low-income households in the newly served areas. A primary document on any of these would improve this page — the “Report an error” link at the foot of this page is how.
Evidence file · public beta 39 addressable claims, 16 source dossiers, one permanent baseline. 31 of them are marked in the text above; the other 8 belong to a passage this page does not print, so they can be read here but not clicked in the article. The file is unreviewed — an AI-assisted structural migration, not a human or expert endorsement. Every revision keeps a permanent copy under revision history.
The text below is copied verbatim from the published record. Each claim names the source dossiers that carry it; caveats stay attached to the claim they qualify.
clm-2e3e132a729b4e1e
quantitative
lost 72% of what it produced
clm-f26981548d84a8dd
quantitative
only 12.61% of its connections had a meter
clm-3d83191d791ab2ea
descriptive
put zoned teams on leak control
clm-a8f84b11aa1beb15
quantitative
Losses fell to 6% by 2008
clm-7390bbf26fa77586
quantitative
bill collection rose from 48% to over 98%
clm-b35ab2f8501261f5
quantitative
From 1994 it began metering every connection, and all were metered by 2001
clm-e14b76e492ffa06c
quantitative
Zone telemetry was in place by 2008
clm-1fe9b8d8b35f43dc
quantitative
the integrated GIS platform came after 2015
clm-d9a31468afc2dea4
quantitative
Non-revenue water fell from 72% in 1993 to 6% in 2008
clm-c18d9057ea1c31d0
quantitative
24-hour supply reaching about 90% of a city of 1.3 million
clm-9295353a5a047a79
descriptive
no state operating subsidy
clm-9fcce6357a4f0d74
quantitative
PPWSA's own reports give 8.75% for 2023, 27.27% for 2024 and 25.55% for 2025
clm-999e92972704c884
descriptive
internal audit is reviewing how the rate is calculated
clm-018aedddd839b91c
descriptive
The turnaround began with a customer census, meters and enforced bills
clm-0bf6d2dd72cb9734
descriptive
zone telemetry followed
clm-01542ef1480fec5a
quantitative
the integrated GIS platform and spot-billing came only after 2015
clm-db2fb78d31685039
quantitative
from a 5.9% low at the end of 2009
clm-e34748ee17968127
quantitative
the operator's own figure climbed to 27.27% in 2024
clm-b1811efcde161374
quantitative
Non-revenue water topped 72% and revenues collected barely covered half of operational costs
clm-a9407c2942b1b04a
quantitative
In 1993 only 3,391 of its 26,881 connections had a meter
clm-6353346196fa7fad
quantitative
they found 13,901 households receiving water but no bill, and 12,980 being billed for water they never received
clm-d0a46e42cdbec7a3
quantitative
an illegal connection cost about $1000 and could be arranged by middlemen linked to the utility's top managers
clm-c97c220ab4eb33f1
quantitative
Metering of every connection began in 1994, and by 2001 all connections were metered
clm-a68256bc8315d8ed
quantitative
By 2008 the distribution network was divided into 41 zones, each equipped with a pressure and flow data transmitter sending online data to find large leaks
clm-6a78b0299093bc51
quantitative
by 2010 process control had been shifted to a SCADA system
clm-d85471893e9f1489
quantitative
a GIS platform combining SCADA with the Sense Water loss-management system came after 2015, and spot-billing from mid-2017
clm-786a9d34339a6315
quantitative
the army initially refused to have any meter installed, and in 1997 the Prime Minister publicly declared that every person and institution had to pay their water bills
clm-9b890f7227d21eee
quantitative
the representatives mostly sold water to the poor at 10 to 20 times the subsidised rate
clm-a65d7b6ba22e7ad4
quantitative
poor households could receive a subsidy of 30, 50, 70 or 100% of the connection fee
clm-aec349cea4c70386
quantitative
connections to poor households rose from 101 in 1999 to 17,657
clm-176ab47fe344a1fb
quantitative
reducing water loss from the current 5.9 percent to 4 percent by 2020
clm-ed3d1654b1bbefae
quantitative
in 2023 the water loss rate was kept in single digits at 8.75%
clm-1b6a30721de74663
quantitative
9.75% in the first quarter of 2024 and 27.17% in the second
clm-f974d31dcb6cdf00
quantitative
27.27% for Phnom Penh and Takmao in 2024 and 25.55% in 2025
clm-b1f963cd2b97b5c0
descriptive
road repair and widening across Phnom Penh is one of the many reasons for leaking pipes
clm-679d764e58ab7b57
descriptive
The board noted that the latest quarterly loss report was not yet a comprehensive one, and internal audit was assessing the methods used to calculate the rate
clm-a5c08d1fd118ccf8
quantitative
15% by 2030 in the chairman's outlook, and 15% by 2027 in the board-approved action plan
clm-5d41c1c7d6a75089
quantitative
The Bakheng plant produced 102,751,262 cubic metres in 2025, with a third phase scheduled for completion in the fourth quarter of 2027
clm-d12cf4877d9baecc
descriptive
The evaluation's charts are marked as drawn from PPWSA records
This layer records what is known about each source, what is still missing, and the exact snapshot this page is rendering. A missing date, archive or hash stays visible as missing.
Awaiting operator and domain-expert review; publication of this revision implies neither.
sha256:fcc2bfedec295b29393146693f64fc77227a730541e8b86e1fd3834c2efab49a
src-8e8f1f55e07367c7Operator's own NRW page: 72% in 1993, 12.61% of connections metered in 1993, meters on all household connections from 1994, Class B to C meter upgrade from 1997, 8 water-loss groups with 63 technical staff, and 8.75% for 2023. Rechecked 2026-09-25: the page still stops at 2023, while the operator's annual reports give 27.27% for 2024 and 25.55% for 2025.
src-868c00d88a1e3bbaReproduces the Stockholm International Water Institute's award announcement: the jury citation (operating without subsidies from the state, 24-hour service, 90% coverage of a city of 1.3 million) plus narrative figures (72% in 1993 to 6% by 2008, collection over 98%) and Ek Sonn Chan's target of 5.9% to 4% by 2020. An award citation, not an audit.
src-d875914f258b27e7Published in IWA's magazine but written by PPWSA's director general, Dr Sim Sitha, so an operator account. NRW above 70% (1993) to 5.9% at end-2009; collection 48% to 99%; a telemetry-run SCADA before 2015, then an integrated GIS with SCADA and Sense Water; spot-billing from mid-2017; 333,288 connections in 2020; dates the third tariff step to May 2017.
src-176b42db6752f691IWA Climate Smart Utilities case story written by PPWSA staff and its director general, Long Naro, so an operator account. Gives 8.93% for 2023 (the annual report says 8.75%) and attributes rising NRW to urbanisation, expansion and an ageing system; Bakheng solar 3.8 MW.
src-f5c283ee4439b583Founding by Royal Decree No. 164NS of 24 March 1960; a 288 km network in poor condition and a 72% loss rate in 1993; 26,881 customers at end-1993 and 365,120 at end-2018.
src-a88c519d262fea8eThe publisher's record of the peer-reviewed evaluation (IJWRD 26(2):157-172, June 2010). The publisher page blocks automated readers; the full text was read in the authors' open copy, listed separately, and the page's claims cite both.
src-b6a75138724f33d2The current capital programme: capacity from 600,000 to about 1,000,000 m³/day; AFD $186.4M, EIB $100M, EU $15M grant, PPWSA $80M; construction started February 2021.
src-8a92186dbc5990b2The full text of the peer-reviewed paper, hosted by the first author's institute (reprint of IJWRD 26(2):157-172, June 2010). Describes itself as 'a summary of an independent evaluation' funded by the Lee Kuan Yew School; its figures and charts are marked 'Source: PPWSA records'. Carries 3,391 of 26,881 connections metered in 1993, all metered by 2001, the 1994 census (13,901 unbilled / 12,980 phantom), 24-hour service by 1995, unaccounted-for water 72% to 6.19%, the failed community-resale scheme, and pro-poor connections 101 (1999) to 17,657.
src-95768af986cd07afADB 'Country Water Action' case study, built largely on interviews with PPWSA management: 72% NRW and revenue covering half of operating costs in 1993; salaries around $20 a month; network divided into 41 zones with online pressure and flow transmitters; 6% NRW, 90% coverage with 24-hour supply, 99.9% collection; no central-government subsidy; 1996 autonomy decree.
src-66cc7a6f19016b0eIUCN Asia study co-authored by Binayak Das with PPWSA's general director Ek Sonn Chan and deputy Chea Visoth, so not independent of the operator. Describes SCADA process control from intake to distribution, a telemeter system detecting leaks by zone, 58 sub-zones with local leak detection, and leak teams paid on the reduction they achieve; NRW 32% in 2000.
src-91089945b788fd6cOperator annual report for 2023: 'water loss rate was kept in single digits 8.75%'. Chairman's letter dated March 2024 (day not printed).
src-0ac7e97d7de9f78dOperator quarterly report filed with the Cambodia Securities Exchange: non-revenue water 9.75% in Q1 2024.
src-d3f7e1c14e92794bOperator quarterly report: non-revenue water 27.17% in Q2 2024, the first figure above 10% in the series; no explanation of the change is given.
src-f774a74959863206Operator annual report for 2024: NRW 27.27% for Phnom Penh and Takmao; names road repair and widening as one of the many reasons for leaking pipes; tariff revised from 1 January 2020.
src-9f875fd5357e6cb0Operator annual report for 2025 (chairman's letter dated 7 April 2026): NRW 25.55%, missing the board's 24% target; targets of 15% by 2030 (chairman) and 15% by 2027 (board action plan); board finds the Q4 NRW report 'not yet a comprehensive NRW report'; internal audit reviewing how the NRW rate is calculated; 1,420 subsidised connections in 2025 in the 30/50/70/100% tiers; Bakheng output 102,751,262 m3 in 2025, Phase 3 due Q4 2027; listing date 18 April 2012; tariff unchanged 2001-2019.
src-6421514aec2539a4Trade/financial press: PPWSA was the first company to list on the Cambodia Securities Exchange, trading from 18 April 2012, with 15% of its shares floated and 85% held by the Ministry of Economy and Finance.
1001 Smart Cities (2026). “PPWSA: Metering Out a 72% Water Loss” — Phnom Penh, Cambodia. The Smart City Atlas. https://1001smartcities.org/projects/phnom-penh-ppwsa-water/ (last verified 2026-09-25). Data: CC BY 4.0.
The underlying data is free to reuse with attribution — see the open data page.