clm-81dfa88990935707
quantitative
SFMTA piloted NextBus arrival predictions on the 22 Fillmore in 1999
Permanent project revision
Revision 2026-08-20-claim-baseline · effective 2026-07-26
Superseded. A later revision exists: 2026-08-21-source-dates →
Created the first structured Evidence Passport with 10 claims and 15 source dossiers.
sha256:f736cff1353fadeb64439cb0fe52f757161ef0045c76f3a7101213a9c3e0c838
add /evidencePassport — Added claim-level evidence mapping and source dossiers.add /revisionHistory — Established a permanent, hash-addressed baseline revision.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-81dfa88990935707
quantitative
SFMTA piloted NextBus arrival predictions on the 22 Fillmore in 1999
clm-d4200859922c3822
quantitative
In 2009 a licensee claimed to own the predictions and got a rider's app pulled from the App Store
clm-fd47eae522a70989
descriptive
the agency insisted the data was public and published a feed that November
clm-d3a3806fea011e30
quantitative
Cubic bought NextBus in 2013, then won the $89M contract to replace it
clm-150b2fb70ee08d16
descriptive
SFMTA states the margin of error is under one minute for predictions of five minutes or less, and under two minutes for ten-minute predictions
clm-e57388b5f772a5a5
descriptive
no independent evaluation of NextMuni's prediction accuracy or of its effect on ridership was found
clm-845c453be2deb71c
quantitative
Over 400 shelter displays lost predictions in June 2022 when AT&T switched off 3G
clm-fe8821e09d0b1ec0
quantitative
San Francisco's predictions only became a public feed after SFMTA wrote ownership into the contract in 2009 and required the vendor to hand over an XML stream
clm-144fddc2c5ac8bdb
descriptive
SFMTA's own deed is revocable and covers timetables only
clm-1e543124178e78a1
descriptive
the real-time feed riders' apps actually use is the region's, not the city's
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:f736cff1353fadeb64439cb0fe52f757161ef0045c76f3a7101213a9c3e0c838
src-3aef36150471d0b5The dispute, sentence-level. NextBus Information Systems (NBIS) told Apple it 'is the sole agent for commercial use of the NextBus real-time prediction data' and held 'exclusive rights to distribution of this data to mobile phones'; Apple pulled the rider-built app Routesy. SFMTA spokesperson Judson True countered that Muni 'owns the data in question and that the public should have access to it'. NBIS was a licensee, distinct from NextBus, Inc., the agency's contractor — the entry keeps them apart.
src-84f2ba513f9ba19aIndependent legal read on the same facts: 'NBIS, like the NYC MTA, appears guilty of copyfraud. They've been unable to produce any proof that they do, in fact, own the data in question.' Records that developer Steven Peterson built Routesy in 2008, that NBIS demanded in June 2009 that it be discontinued, and that his attorney got Apple to restore it once SFMTA clarified ownership.
src-49b1901f94e7f737The resolution and the contract language the entry rests on: the renegotiated contract 'expressly declared that the MTA owns the data' and obliged the vendor that 'Contractor shall provide necessary technological assistance in placing an XML data stream (via weblink or other SFMTA approved data conveyance)'. Ends a five-month dispute. This is the date a public developer feed existed — ten years after the pilot, and separate from the pilot, the citywide contract and the street displays.
src-a45b8c03be25dc8eOperator chronology and self-criticism: 'When the SFMTA first partnered with NextBus more than twenty years ago to create NextMuni, the real-time transit vehicle arrival predictions system was state-of-the-art.' Names the failure modes the replacement targets — inaccurate predictions near the start of routes, and 'ghost buses' that are predicted and never arrive.
src-de1cc422f4c752f4Currency check, July 2026: status still 'Implementation / Construction'; the 2023-2024 phase 'System Enhancements and MuniMobile App' is marked Completed; capital cost line $25.4 million (the SFMTA capital share, not the contract ceiling). Promises 'a more sophisticated and accurate vehicle arrival prediction algorithm' but publishes no accuracy target or measurement, which is why none is claimed here.
src-1bed567e01f2a4baThe replacement contract, vendor-announced: SFMTA contracted Cubic Transportation Systems 'in an amount not to exceed $88,980,877, for an initial term of six years with two optional five-year terms'. The same amount appears in the SFMTA board calendar item of 16 June 2020 (PDF too large to render in this environment; figure taken from the indexed text and matched against the vendor release).
src-ed5a6eccc12f3d4eThe vendor chain: NextBus, Inc., founded 1997 in Emeryville, passed through Grey Island Systems International and Webtech Wireless before Cubic 'acquired all of the outstanding capital stock of NextBus, Inc. from Webtech Wireless, Inc.' on 24 January 2013 for CAD$20.75 million. NextBus was then in over 100 North American agencies. Cubic later folded the brand into Umo and moved the public API endpoints off webservices.nextbus.com to retro.umoiq.com.
src-38f90720a202bcf1The licence position, verbatim: 'nonexclusive, limited, and revocable'; mandatory notice on derivative works — 'Reproduced with permission granted by the City and County of San Francisco…'; 'The License granted hereunder does not include any right to sublicense.'; SFMTA 'reserves the right to alter or cease to provide the Data at any time without prior notice'; supplied 'as is' with no warranty of accuracy. And the scope limit that matters: 'This is not real-time data, only scheduled service.'
src-8aca195a7a160a44Where the real-time feed actually lives today: the regional 511 portal, not the city's. Carries GTFS, GTFS-Realtime (Trip Updates, Vehicle Positions, Service Alerts) and SIRI/NeTEx APIs, lists SFMTA among operators, requires a registered API token, defaults to '60 requests per 3600 seconds', and applies its own data agreement on top of any operator terms.
src-e52d92b7422a41e7Counterweight, sentence-level: 'over 400 NextMuni displays that are currently utilizing AT&T 3G modems will no longer be able to display real-time Muni vehicle arrival predictions'. The street-level layer of a 23-year-old system was switched off by a carrier, not by the agency. Replacement: 'over 700 new Liquid Crystal Display (LCD) displays'.
src-b4f40956f85d259bRollout progress: over 435 of 800+ planned LCDs installed, replacing the old NextBus signs, 'with an emphasis on underserved neighborhoods'. New displays add crowding data from automatic passenger counters and short-term route changes.
src-73a80654ebf38adcThe only accuracy figures found, and they are the operator's own: 'The margin of error overall is less than one minute for predictions of five-minutes or less. For 10-minute predictions, the margin of error is less than two minutes.' The page carries no date, sample or method, and adds that the times 'are simply predictions'. A circulating counter-figure — predictions wrong close to half the time at commute hours, credited to a study by the app company Swyft — could not be verified sentence-level (the carrying SF Examiner article returned HTTP 429 repeatedly) and is deliberately not used.
src-70cbdb689a719571Transportation Research Record, 2018. A stated-preference survey of 5,856 Muni customers in English, Spanish and Chinese, run by SFMTA to design the replacement system: facing a 20-minute wait, respondents choosing Muni rose 'from 45 percent with basic information… to as much as 72 percent (with a smartphone app) or 82 percent', and with countdown signs offering alternative routes 'differences among income groups virtually disappeared'. Peer-reviewed but agency-authored, and it measures hypothetical responses to proposed designs, not the deployed system — so it does not lift this entry's evidence grade.
src-f72a50b6e22bda30Independent corroboration of the start year — the project 'replaces the 23-year-old NextBus technology (launched 1999)' — and of the handover: SFMTA CTO Lisa Walton called the November 2022 switch 'a soft launch of the new software', phased through 2024, with roughly 800 solar-powered signs for locations without power.
src-6ae21ee7ef204ce0Why this entry claims no superlative. London Transport trialled Passenger Information at Bus Stops on Route 18 in 1992 and installed the first Countdown signs in autumn 1992 — street-level arrival displays seven years before San Francisco's pilot; the 1993 conference paper independently fixes the system before 1994. A US research precedent also exists: MyBus, built at the University of Washington by Daniel Dailey in the mid-1990s, which was not verified sentence-level here and is therefore only noted, not asserted. San Francisco is early and consequential, not first — and it is by some years the oldest real-time transit feed in this atlas.