Street Bump Road Defect App
Boston's New Urban Mechanics office started Street Bump in 2011 and put the public version out in 2012: an app that read a driving phone's accelerometer and GPS and filed road defects without anyone tapping anything. Its biggest finding was not potholes but sunken manhole covers. The city ended the project in 2014.
Completed
📅 2011–2014
✔ Editor-reviewed
📊 Impact
The city reports that sunken manhole covers caused about four times more bumps than potholes, that it fixed 1,250 of the worst covers by working with the utilities that own them, and that testing held false positives under 10%. Boston never published how many reports the app generated, how many were confirmed, or whether repairs got faster.
🎓 Lesson
Kate Crawford's 2013 Harvard Business Review essay made Street Bump the standard example of a sensing system's 'signal problem': a phone-and-car sensor samples people who own phones and cars. That was an argument from smartphone ownership, not a measurement of Street Bump's own reports — Boston never published a neighbourhood breakdown, so the most-cited critique in civic sensing was never tested against the data it describes.
Evidence🏛️ Official
Scale🏙️ City-wide
FundingCity of Boston; the algorithm competition's $25,000 prize pool was donated by Liberty Mutual through InnoCentive
Sources
↗ City of Boston — Street Bump (Mayor's Office of New Urban Mechanics)The operator's own page and the authority for the load-bearing facts. Status, sentence-level: 'The involvement of New Urban Mechanics in this project was from 2011 to 2014. The project is not currently active.' Also the three phases (2011 alpha, the public algorithm competition, the 2012 accelerometer-and-GPS release), the finding 'Residents most frequently reported problems about potholes, but the biggest cause of bumps is sunk manhole covers' at roughly four times the rate, the 1,250 covers fixed with the utilities, 'false positives under 10%', and the partner list: Fabio Carrera, Red Fish Group, IDEO, Connected Bits, InnoCentive and Boston University researchers. No statement of why the project ended appears anywhere.
↗ CNN — Street Bump app detects potholes, tells city officials (16 February 2012)Independent account of the mechanism and the aggregation rule: the phone sends accelerometer data to a server that combines many phones, and 'If at least three people hit a bump in the same spot, the system recognizes it as a pothole.' Names Nigel Jacob of New Urban Mechanics as project lead and states the city's hope of replacing survey trucks with a real-time map.
↗ WBUR — Hit A Pothole? Tell Boston With New Smartphone App (24 April 2012)Local public-radio coverage of the 2012 public release, used as the second source for the launch year.
↗ MIT Technology Review — Road Repair via Crowdsourcing (13 May 2011)Contemporary account of the algorithm competition: after the first build could not tell potholes from other bumps and produced too many false positives, the city turned to InnoCentive, with $25,000 in prize money donated by Liberty Mutual. The contest drew more than 700 solvers and 19 submitted solutions.
↗ Aboufadel, Marculis et al. — Wavelet-Kruskal Solution to the InnoCentive Boston Pothole Challenge (Grand Valley State University)A winning entry, from the winners' own institutional record: three winning teams were announced in December, two of them awarded $9,000 each. The method was wavelet analysis and Kruskal's algorithm — signal processing, not machine learning, which is why this entry carries ai: false.
↗ Boston Globe — App shows jarring role of cast-metal covers in Boston (16 December 2012)Contemporary independent press on the manhole-cover finding. The Globe paywall blocks automated fetch; the outlet, date and subject were confirmed by search, and the substance is carried sentence-level by the city page above.
↗ Kate Crawford — The Hidden Biases in Big Data, Harvard Business Review (1 April 2013)The origin of the critique, and the reason for the lesson's careful wording. This is a business-magazine essay, NOT a peer-reviewed paper: Crawford writes that 'StreetBump has a signal problem' because lower-income and older residents are less likely to own smartphones. The argument is drawn from smartphone-ownership demographics; neither this piece nor any source found analyses Street Bump's own reports by neighbourhood. HBR is paywalled to automated fetch; the passage is reproduced by the textbook below.
↗ Barocas, Hardt and Narayanan — Fairness and Machine Learning (Cambridge University Press), IntroductionCarries Crawford's Street Bump passage verbatim and shows how the case entered the algorithmic-fairness canon: 'Consider Street Bump, a project by the city of Boston to crowdsource data on potholes' … the data reflect 'the patterns of smartphone ownership, which are higher in wealthier parts of the city compared to lower-income areas and areas with large elderly populations.' The textbook presents it as an observation about the method, with no measurement of the project's data.
↗ US Federal Trade Commission — Big Data: A Tool for Inclusion or Exclusion? (January 2016)The closest thing found to a documented Boston response: the FTC uses Street Bump as its worked example of a 'data desert', stating that once the team recognised that lower-income residents were less likely to carry smartphones, it recognised its data was not representative of road conditions across Boston. The report PDF resisted text extraction, so this is verified at search level only and is not quoted as the city's own words anywhere in this entry.
↗ GovTech — Boston Testing App for Auto-Detecting PotholesConfirms that the pre-release version was run on city inspectors' vehicles before public release, and describes the pipeline: data to Connected Bits' servers, likely problems submitted to the city through Open311 and classified as potholes to fix or known obstacles such as speed bumps.
↗ City of Boston Analytics Team — Automated Pothole Detection (Mercedes-Benz USA pilot, 2023)What came after, from the same offices. The Analytics Team and New Urban Mechanics tested anonymised Mercedes-Benz vehicle sensor data against 311 reports and found it 'particularly helpful for identifying potential potholes on larger streets with higher speed limits, where 311 reports are relatively rare' — passive sensing used explicitly to cover what the complaint channel misses. It also repeated Street Bump's classification problem: 'other types of irregularities in the road were getting picked up, such as depressions near utility covers.' The page does not mention Street Bump, so no lineage is claimed as fact.
📎 Cite this project
1001 Smart Cities (2026). “Street Bump Road Defect App” — Boston, United States. The Smart City Atlas. https://1001smartcities.org/projects/boston-street-bump/ (last verified 2026-07-26). Data: CC BY 4.0.
The underlying data is free to reuse with attribution — see the open data page.
🔗 Related projects
🇺🇸 CleanStat Street Cleanliness Index · Los Angeles
Boston let driving phones file the reports and became the textbook case of sampling bias; LA drove every block itself from 2016.
🇺🇸 BOS:311 and Boston's Service-Request Record · Boston
Both ran out of New Urban Mechanics: the app filed bumps nobody reported; 311 waits for residents, 80% of them within two blocks of home.
🇨🇦 Montreal's GPS-tracked snow clearing and Info-Neige alerts · Montreal
Montreal is the other half of the same idea: GPS on every snowplough, yet the resident-facing status is still typed in by the crews.
🇺🇸 CityScore Performance Dashboard · Boston
Another verified project from Boston.
→ Open this project in the interactive atlas · → 🏛️ Governance category · 🚩 Report an error