Park Sight

See where and when parking tickets are issued in Philadelphia, and where they are likely to be next. Built on 9.4 million city records, 2012–2017.

Heatmap explorer

Map of ticket density for any month, with a time-of-day slider, playback over six years, and a zoom-in view of every individual ticket.

Open the map →

Tickets over time

Monthly and yearly ticket volume, showing the steady rise from 1.5M tickets a year to 1.72M in 2017.

See the chart →

Hotspot forecast

Pick a date and time window to rank the street blocks expected to see the most tickets, with CSV export. Backtested on held-out 2017 days.

Forecast a day →

About the data. Source: City of Philadelphia parking violations via OpenDataPhilly. The city's 2018-to-present data was never published, so everything here covers 2012–2017. About 8% of tickets have no coordinates and don't appear on maps. Forecasts for later dates assume the 2012–2017 pattern still holds. See the forecast backtest for accuracy.

Why it matters

Blocked bike lanes and bus lanes are a safety and transit problem, and parking enforcement staff are limited. Park Sight shows the Philadelphia Parking Authority (PPA) where enforcement is concentrated hour by hour, and where it is missing.

Our approach

Where patrols are today. The hotspot forecast ranks street blocks by expected tickets for any date and 4-hour window.
Where they aren't (next step). Overlay bike lanes, bus lanes, school zones and crash corridors to find safety-relevant blocks the ticket data under-covers. That is how a patrol scope widens instead of just reinforcing itself.

How well it works

9.4Mtickets, 2012–2017

10,255street blocks modeled

~25×better than random: top 50 blocks capture 15.8% of a day's tickets vs 0.6%

~75%of the best possible hindsight score (21.2%)

Trained on 2012–2016, tested on 123 unseen days of 2017. Weekday and time-of-day awareness both help, most clearly on weekends. Details in the backtest.

Supporting research

Reducing double parking can raise traffic speeds roughly 10–15% and cut delay roughly 15–20% in simulation (arXiv). These are traffic-flow results, not crash counts.
Illegal parking has been linked to collisions, and a New York study cited by Toronto found 88% of cyclists swerve into traffic when a bike lane is blocked (Toronto report).
Researchers note the safety effects of double parking are not well quantified (UMass Amherst). We make no claim that enforcement is proven to prevent crashes.
The PPA runs a Vision Zero bike lane enforcement unit and bus-lane cameras with SEPTA, where a pilot flagged over 36,000 violations in about 70 days (CBS Philadelphia). A hotspot forecast tells human patrols where to go. Cameras cover what patrols can't.
Concentrating resources where problems cluster is among the best-supported strategies in crime policing (College of Policing). We apply the same logic to parking. It has not been tested for parking specifically.

Honest limits

Data ends in 2017. The city never published later data. A records request to the PPA is drafted, and current data would replace the old data.
Feedback loop. Tickets show where enforcement happened, not pure parking demand. Patrolling only the top-ranked blocks would make them rank even higher (Lum and Isaac). The coverage-gap layer is our mitigation.
No holidays, weather, events or staffing inputs yet.

Our ask: a data partnership with the PPA.