clm-dd43c3c411b2c9de
quantitative
picked Aclara's fixed radio network in 2009
SFPUC wired 175,000 water meters for hourly readings in 2010, built leak alerts an outside evaluation says work — and then the transmitters aged out and 8,000 households stopped getting bills. The full life cycle of smart-city hardware, in one programme.
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.
37 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
✎ Updated 20 Aug 2026 Publication date recorded for 13 sources. · 2 revisions in all What changed →
Part of 1001 Smart Cities — 1,167 documented projects · 🏙️ 9 more in San Francisco · 💧 Water guide · 📡 Data & IoT projects
The radio is smaller than a fist. It sits in a cast-iron box under the pavement, wired to the meter beneath it, and it talks for about a tenth of a second at a time. San Francisco began fitting them in June 2010, one per address, a decade before most American utilities thought the money worth spending. They did what the brochure said for roughly twelve years. Then they started to fail, faster than crews could replace them, and by September 2024 about 8,200 customers had quietly stopped receiving bills.
Until 2010, most residential meters in San Francisco were read by a person walking the block once every two months. In November 2009 the San Francisco Public Utilities Commission picked Aclara's STAR fixed radio network, and on 7 June 2010 it began what was announced as a two-year, $50 million replacement of more than 175,000 meters, funded by a 91-cent charge on every monthly water bill. The two years slipped — the Alliance for Water Efficiency dates the installation as "primarily occurring from 2010 to 2013", and a final-completion line was still sitting in the capital budget in FY2021 — but the coverage claim held: SFPUC says transmitters are now in place at nearly all of the city's 178,000 accounts.
The design is worth pausing on, because it explains both the system's endurance and its ending. Each meter transmission unit takes a reading every hour and sends its batch four times a day, in radio bursts of less than a tenth of a second, to one of 81 collectors around the city. There is no network listening for a wake-up call: the transmitter fires on a timer and goes back to sleep. That transmit-only discipline is what let the vendor's 2008 specification sheet promise a "20-year permanent battery" — a number that becomes important later in this story.
| When | What happened | Who says so |
|---|---|---|
| Nov 2009 | SFPUC selects Aclara's STAR fixed radio network for a citywide rollout | Trade press / vendor |
| Jun 2010 | Installation begins — announced as a two-year, $50M effort funded by a 91-cent monthly bill charge | Press |
| 2013 | Installation "primarily" complete; over 99% of ~180,000 retail connections on the system | Independent (AWE) |
| May 2014 | My Account portal opens hourly and daily usage to customers — launched mid-drought, four years after installation began | Operator / press |
| 2015 | Leak notification piloted as weekly postcards, driven by a weekly Aclara continuous-usage report | Independent (AWE) / trade press |
| 2017 | The automated Leak Alert programme replaces the postcards | Independent (AWE) |
| Aug 2021 | Alerts extended beyond homes to commercial and large multi-family accounts; SFPUC estimates 47M gallons saved in the prior fiscal year | Operator |
| Mar 2023 | The Alliance for Water Efficiency publishes the programme's only outside statistical evaluation | Independent |
| Sep 2024 | The billing failure becomes public: ~8,200 customers unbilled for months, some over a year, as ageing transmitters stop reporting | Press / operator |
| 2026 | SFPUC's delayed-bills FAQ still speaks in the present tense; no successor contract has appeared in the Commission's public resolution logs | Operator, by omission |
The point of hourly data is not the bill — a bimonthly walk-by could produce that. The point is the shape of the curve. A household that never sleeps is a household with a leak, and SFPUC's alert fires when a small account runs 7.5 gallons an hour or more for 48 hours without a break. Large buildings get subtler screens: night-time consumption between 1 a.m. and 4 a.m. compared against the account's own 90-day average, because a hotel that never sleeps is normal and a school that never sleeps is not.
Nobody signs up for any of this — it is a blanket courtesy across every transmitting meter. And the survey data SFPUC has collected since August 2020 is the best argument for it: of 2,138 responding households, 77 percent did not know about the water use until the notice arrived. More than half traced it to a leaky toilet; 45 percent fixed the problem themselves without a plumber. The follow-through is patient rather than forceful — a second notice after two weeks, a third after ten, and after 100 days an inspector hangs a final paper notice on the door. What never follows is enforcement: no fine, no shut-off, no obligation. San Francisco's alert is information, offered politely, with the water still running.
There is an honest gap in the delivery layer, and SFPUC's own case study names it: the utility has a phone number or e-mail address for only about half of its customers. Everyone else gets the slow path — a letter.
Most leak-alert programmes in this atlas rest entirely on the operator's own arithmetic. San Francisco's is the unusual one: in March 2023 the Alliance for Water Efficiency published a four-utility statistical evaluation with SFPUC as a case study, comparing mean leak volume per meter before and after each programme launched.
Single-family homes also showed a statistically significant reduction (about 0.63 gallons per meter per day); the study reports no clean percentage for SFPUC's single-family programme alone. Pre/post design with a nine-month baseline, no control group; SFPUC part-funded its own case-study participation.
The most telling row is the one that failed. Alerts to the city government's own irrigation meters — parks, medians, public grounds — produced no significant effect, while identical alerts to private irrigation accounts cut leak volume in half. The difference was not the technology: private customers were notified by an automated daily e-mail, while the city's own departments were chased manually, "depending on the availability of water department staff". The utility that automated leak detection for everyone else had not automated the follow-up to itself. SFPUC has since fixed this — but as a finding about how these systems really work, it is hard to beat: the sensor is never the bottleneck; the person the message reaches is.
One caution against over-reading: the AWE study is a pre/post comparison, not a controlled trial, and SFPUC was one of four utilities that co-funded its own case study. The nearest gold-standard evidence for the mechanism — a randomized controlled trial published in Environmental Research Letters in 2024, which attributed 92.8 percent of measured conservation to leak detection and repair — ran across the bay at East Bay MUD, a different utility. It supports the genre, not this programme.
Every fixed-network meter system carries a countdown that nobody advertises on launch day. The vendor's specification sheet promised a 20-year battery; SFPUC's own capital plan budgets $40.2 million to replace "the entire AWMP System by the end of its 20-year useful life (ending in 2031)". The failures did not wait for the schedule. By 2024, transmitters installed in 2010–2012 were falling silent in numbers the repair crews could not match, and the meters beneath them — still measuring correctly — had no way to tell the billing system what they saw.
Staffing and equipment shortages left us unable to keep up with repairing our aging automated metering equipment. The situation was exacerbated by supply chain delays.
— SFPUC, Delayed Water and Sewer Bills FAQ
The public found out through the mailbox, not a press release. In September 2024, local outlet 48hills reported that some 8,000 households had not received a water bill in months; television news put the count at about 8,200, some unbilled for over a year. The catch-up bills were real money — at a typical $140 a month for a single-family home, ten silent months meant a four-figure demand. SFPUC waived interest and late fees and offered payment plans, but the consumption itself stayed owed: the courtesy that never enforced a leak alert also never forgave a bill.
What happens next is, as of mid-2026, genuinely unwritten. SFPUC's FAQ says it is "evaluating and upgrading" the transponder system, and its job postings name an AMI Lifecycle Management Project — but no successor contract, vendor, cost or schedule has appeared in the Commission's public resolution logs. A city that spent $50 million learning what hourly water data is worth now gets to decide, in public, what the second generation is worth.
An independent 2023 evaluation by the Alliance for Water Efficiency found the leak alerts cut mean leak volume per meter by a statistically significant margin for single-family homes, by 18% for multi-family buildings and by 51% on non-municipal irrigation meters — while alerts to the city's own municipal meters, still handled manually at the time, showed no significant effect. SFPUC itself estimated 47 million gallons saved in the fiscal year before August 2021, and in its surveys 77% of notified customers had not known they had a leak.
A meter that reports hourly can still fail to bill. The vendor's data sheet promised a 20-year battery, and SFPUC's capital plan schedules replacement of the system by the end of its 20-year useful life in 2031 — yet by September 2024 about 8,200 customers had gone months, some more than a year, without a bill, because the utility could not keep up with repairing its ageing transmitters. The alerts, meanwhile, remain a courtesy with nothing enforced behind them.
Two clocks, kept apart: when this entry’s sources were published, and what has happened to its record in the atlas. Documented span: since 2010. 10 of 13 sources carry a publication date, from 2008 to 24 September 2024.
Thirteen sources carry this entry: SFPUC's own programme pages, capital plan and FAQ, the AWE evaluation, the vendor's specification sheet, and the press accounts that broke the billing story. Four things the record honestly cannot answer: who physically installed the meters (no contractor is named in any reachable document), the final as-built cost (press figures range from $50M in 2010 to $56M by 2014, with no primary cost document public), when the original rollout was formally declared complete (no announcement exists), and what the replacement will cost. Anyone who finds a primary source for any of these improves this page — the “Report an error” link at the foot of this page is how.
Evidence Passport · public beta 43 addressable claims, 13 source dossiers, one permanent baseline. 37 of them are marked in the text above; the other 6 belong to a passage this page does not print, so they can be read here but not clicked in the article. The passport 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-dd43c3c411b2c9de
quantitative
picked Aclara's fixed radio network in 2009
clm-b7ac55b8e0995f57
quantitative
began replacing San Francisco's roughly 175,000 water meters in June 2010
clm-18482db4b092703d
quantitative
Transmitters now cover more than 96% of accounts
clm-bb1afc5f7cb1c29e
quantitative
sending hourly readings to 81 collectors
clm-73127f16c6fc714d
quantitative
a courtesy leak alert flags 48 hours of constant flow
clm-e728af0bf02ba69f
descriptive
Ageing transmitters have since left thousands unbilled
clm-30d6a16775d4780f
descriptive
cut mean leak volume per meter by a statistically significant margin for single-family homes
clm-35b19c7ef69f5ded
quantitative
by 18% for multi-family buildings
clm-060bdddbaff03bc6
quantitative
by 51% on non-municipal irrigation meters
clm-9518e21f7be4da53
descriptive
alerts to the city's own municipal meters, still handled manually at the time, showed no significant effect
clm-c5756bfa2ff2df74
quantitative
SFPUC itself estimated 47 million gallons saved in the fiscal year before August 2021
clm-8ca4743dd6d51d2d
quantitative
77% of notified customers had not known they had a leak
clm-f69a46cb398ca6c4
quantitative
The vendor's data sheet promised a 20-year battery
clm-3267d66073270131
quantitative
schedules replacement of the system by the end of its 20-year useful life in 2031
clm-d24450b2fa17c9ea
quantitative
by September 2024 about 8,200 customers had gone months, some more than a year, without a bill
clm-46dd1a498ff91f72
descriptive
the utility could not keep up with repairing its ageing transmitters
clm-0b04a6c6fccab9d2
descriptive
remain a courtesy with nothing enforced behind them
clm-ac80ae8f692e2e2f
quantitative
by September 2024 about 8,200 customers had quietly stopped receiving bills
clm-2a61776afae44118
descriptive
most residential meters in San Francisco were read by a person walking the block once every two months
clm-3252d379f914577f
quantitative
In November 2009 the San Francisco Public Utilities Commission picked Aclara's STAR fixed radio network
clm-35c58caff4b6cc1a
quantitative
on 7 June 2010 it began what was announced as a two-year, $50 million replacement of more than 175,000 meters
clm-fa89a71a6479879c
quantitative
funded by a 91-cent charge on every monthly water bill
clm-c174fd8c97910419
quantitative
the Alliance for Water Efficiency dates the installation as "primarily occurring from 2010 to 2013"
clm-6c5b7f1ba86b2ecc
quantitative
a final-completion line was still sitting in the capital budget in FY2021
clm-37d496ac573523ba
quantitative
transmitters are now in place at nearly all of the city's 178,000 accounts
clm-46769ad28bd44458
quantitative
Each meter transmission unit takes a reading every hour and sends its batch four times a day, in radio bursts of less than a tenth of a second, to one of 81 collectors around the city
clm-49b929e8ddca3dd5
quantitative
"20-year permanent battery"
clm-202d33c9c9c25ae4
quantitative
alert fires when a small account runs 7.5 gallons an hour or more for 48 hours without a break
clm-91887ec47240c5f2
quantitative
night-time consumption between 1 a.m. and 4 a.m. compared against the account's own 90-day average
clm-d6c557a68628d709
quantitative
77 percent did not know about the water use until the notice arrived
clm-20184d496af14a7e
quantitative
More than half traced it to a leaky toilet; 45 percent fixed the problem themselves without a plumber
clm-79b8039a596c31fa
quantitative
a second notice after two weeks, a third after ten, and after 100 days an inspector hangs a final paper notice on the door
clm-59b5fe2bb679b6fb
descriptive
the utility has a phone number or e-mail address for only about half of its customers
clm-172e7856f32940de
quantitative
in March 2023 the Alliance for Water Efficiency published a four-utility statistical evaluation with SFPUC as a case study
clm-820eeb3f466ccfc8
descriptive
identical alerts to private irrigation accounts cut leak volume in half
clm-e4932834068ce5e7
descriptive
"depending on the availability of water department staff"
clm-ae2859bd7788ea8c
quantitative
attributed 92.8 percent of measured conservation to leak detection and repair
clm-847d5fc46e37a82b
quantitative
budgets $40.2 million to replace "the entire AWMP System by the end of its 20-year useful life (ending in 2031)"
clm-99a8dbfaefc73ed0
quantitative
local outlet 48hills reported that some 8,000 households had not received a water bill in months
clm-8cd5b84fc8a6c95b
quantitative
television news put the count at about 8,200, some unbilled for over a year
clm-98f79c5448c62190
quantitative
at a typical $140 a month for a single-family home
clm-4c1c4e53f6df19d4
descriptive
SFPUC waived interest and late fees and offered payment plans, but the consumption itself stayed owed
clm-7f9bc34b1acd5a26
descriptive
says it is "evaluating and upgrading" the transponder system
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:c1972b1931b16cc6cc29ce6c0b490f35d9e158e9a9ada283bcdacb99cd802de4
src-05d56ac19f794358The operator's own description of the system. Supports: 'Meter transmission units are now in place for nearly all of San Francisco's 178,000 water accounts'; the Aclara Fixed Network AMI STAR System reading 'more than 96% of San Francisco's water accounts'; the MTU collecting a reading each hour and sending to a collector 'Every six hours'; 'A total of 81 Data Collectors are located throughout the city'; and the My Account data going live in Spring 2014 with downloadable daily and monthly reports.
src-a77ce079bd7d6b86Currency check and the mechanics. A 'courtesy Leak Alert Program' that 'reviews hourly water use for all meters that transmit data to us electronically', with no sign-up needed; alerts after '48 hours of continuous water use of 7.5 gallons an hour or more' at small accounts, and on overnight use 100% or 300% above the 90-day average at large ones. Gives the current operating figure: 'over 170,000 meters serving 165,000-plus customers'. Confirms My Account is the live customer portal in 2026.
src-926fe954f024ded3The only quantified result found, and it is self-reported: 'In the last fiscal year, the Leak Alert program saved SFPUC customers an estimated 47 million gallons of water.' Also dates the extension of alerts beyond homes to 'commercial and industrial properties and larger multi-family buildings', and gives the commercial thresholds (increases of 50 percent or more in use over 90 days, plus nightly spikes).
src-068f65ccc40d3705Anchors the start and the money: the SFPUC 'on June 7 began its automated water meter program, a two-year effort'; 'San Francisco started replacing its 175,000-plus water meters in June'; and 'The $50 million program is being funded by a 91-cent fee that's being tacked onto ratepayers' monthly bills.' The cost and the funding mechanism rest on this press account only — no SFPUC or Controller cost document was reachable. Also records that the online-viewing component was expected to lag the installation, which it did by four years.
src-3a2f4d31d443496dDates the vendor selection to November 2009 and describes the STAR Network: low-frequency radio, hourly readings sent four times a day, replacing manual reads taken every two months residentially. Carries the claim that SFPUC 'will be the first major California water utility to install Advanced Metering Infrastructure (AMI) technology to track water consumption' — a vendor-announcement claim in trade press, deliberately not used in the entry.
src-08dc890a504c86b0The utility's own account of the failure. 'A technical problem led to some customers not receiving a water and sewer bill for several months. The problem was caused by certain transponders on water meters' — the meters read correctly but 'the transponders were not sending that data electronically to our billing system.' The admission: 'Staffing and equipment shortages left us unable to keep up with repairing our aging automated metering equipment. The situation was exacerbated by supply chain delays.' Also states SFPUC is 'in the process of evaluating and upgrading our meter transponder system', which is the only evidence of a second generation — no contract, count or schedule is given.
src-4e6e4249e6d8ce0bScales the billing failure: about 8,200 residents affected, some three months without a bill and others a year or more, with SFPUC saying customers are not charged for the repair but remain liable for the service received. Press account; SFPUC has not published a customer count of its own.
src-32de3bffb656d438The evaluation that carries this entry's grade — the only outside statistical assessment of the programme. Dates the rollout ('implementation primarily occurring from 2010 to 2013', 'Over 99 percent of its approximately 180,000 retail service connections'), the portal (2014; 'about 53 percent of SFPUC's customers were registered' by spring 2022), the postcard pilot (2015) and the automated programme (2017). Results: SFPUC's single-family programme is one of 'three of the four' with statistically significant reductions; multi-family 'registered an 18 percent reduction'; non-municipal irrigation meters 'a statistically significant 51 percent reduction'; 'Notifications to municipal meters did not result in a statistically significant difference.' Survey (n=2,138): '77 percent were not aware of the usage until they received the notice.' Disclosed limits: pre/post comparison with a nine-month pre-period, no control group, and SFPUC is one of four utility partners that 'provided funding to serve as case studies' — the reason this grades independent, not peer-reviewed.
src-079a5e52f9345467The utility's own replacement arithmetic: 'Local Water – Automated Meter Reading System: $40.2 million. This program provides funding for the ongoing Automated Water Meter Program (AWMP) including meter renewal, replacement, automation, and replacement planning for the entire AWMP System by the end of its 20-year useful life (ending in 2031).' The transmitters started failing at scale seven years before that end date. An earlier scenario line ('AWMP Completion (FY21)') also shows final installation was still a budget item in FY2021.
src-4e4219ba1d9b4c34Vendor material, used deliberately as the claim the failure is measured against, not as support: 'Contains a 20-year permanent battery.' Also documents the design that made the battery last: 'STAR Network MTUs are capable of one- or two-way communications' but transmit on a timer 'over FCC-licensed, 450- to 470-MHz radio frequencies' — in San Francisco the deployment operates transmit-only, per SFPUC's own description of six-hourly sub-second bursts. Mirrored on a distributor site; the vendor's original page is gone.
src-45982861eb30666bThe report that broke the billing story, five days before the TV coverage: 'Some 8,000 households in San Francisco haven't received a water bill in months.' Scales the hardship — at typical bills of roughly $50/month for an apartment and $140/month for a single-family home, 'ten months without a bill could mean a sudden payment of more than $1,000' — and carries the general manager's response that SFPUC 'plans on accommodating' hardship cases via payment plans and the 25-40% low-income discount.
src-aad1e0fd7076a98bTrade press quoting the programme's manager mid-drought, when the alerts were still weekly postcards: 'The Aclara report tells us which accounts have exhibited continuous usage every hour over a three-day reporting period each week' and 'The utility reaches out to those who show up on the report by sending weekly postcards.' Frames the drought context: utilities scrambling 'to meet government mandates for a statewide 25 percent reduction in water usage.'
src-8e38aaa134894c7dContext only — this randomized controlled trial ran at East Bay Municipal Utility District across the bay, not at SFPUC, and must never be read as evidence for this entry. It is cited because it is the nearest peer-reviewed test of the same mechanism: '92.8% of the estimated water conservation is attributable to leak detection and repair.' It shows what the gold-standard evaluation SFPUC has never had would look like.
1001 Smart Cities (2026). “SFPUC Automated Water Meter Programme” — San Francisco, United States. The Smart City Atlas. https://1001smartcities.org/projects/sf-automated-water-meters/ (last verified 2026-08-12). Data: CC BY 4.0.
The underlying data is free to reuse with attribution — see the open data page.