Mapbox Cities
Mentorship program for cities to get smarter
at making data-driven decisions
Open by default
Urban challenges
New technologies
Improving cities faster
Return on Investment (RoI)
Organizations can expect increased
RoI when investing in open source
technology (World Bank)
Traditional
“closed-by-default”
Open
Source
at least 200%
Open Source
for Cities:
Get better, faster!
Copy what
works.
Repeat.
Alarming rise in traffic fatalities
2014
Traffic fatalities in the
United States
2015
Sharp increase of traffic
fatalities by 7.2%
A wave of
‘Vision Zero’
initiatives...
35,09232,675
Alarming rise in traffic fatalities
2014
Traffic fatalities in the
United States
2015
Sharp increase of traffic
fatalities by 7.2%
35,09232,675 37,461
2016
Number
continued to rise...
?
Vision Zero “No loss of life is
acceptable.”
Vision Zero DC
Data - driven decision making
www.dcvisionzero.com
Washington, DC
Census data on race and ethnicity in DC
Goal: Act faster
OPEN
DATA
POLICY ACTION
Time
Timeline 2017
Dec 2016
Start conversations
with DDOT
Kicking off
In-person meeting
in DC
Smart Cities
Week DC
Deadline to present some
insights to the public
Open Belgium
Sharing our take
aways
2016 2017 April - Aug - Oct Nov Dec 2018 - Mar
How can we
better prioritize
where measures are needed?
DC’s crash data
150,000+ entries
45+ attributes
Data quality
Cause of errors obscured by
complex internal processes
Format changes
Split data into two separate sets
that couldn’t be joined b/c of #1
Assumption 1
More vehicles or pedestrians
= more opportunities for incidents
Assumption 2
Higher speeds = more crashes
Assumption 3
More shops, restaurants & schools
= higher frequency of crashes
CC BY-ND 2.0 by Daniel M. Hendricks | Flickr
Data available
opendata.dc.gov
Crash data (before 2017)
Census data
Intersection data
DDOT + Howard University Traffic Data Center
Traffic counts
Mapbox
Mobile sensor data (speeding)
Modeling Collision Frequency
Various conditions
Traffic counts
Employment data census block
Intersections
School locations
Mapbox speed data*
Density of
crashes
Lively urban streets.
More accidents.
Intersection
density
matters.
CC BY-ND 2.0 by Sonara Arnav | Flickr
Higher speeds
More incidents
= unrelated
School locations
Crash frequency
= unrelated
Why took it so long?
Dec 2016
Start conversations
with DDOT
Talking legalese
Finally we agreed that no
contract is ok, since only
using open data.
Testing datasets
With distributed team, no clear
guidance from the DDOT team
Smart Cities
Week DC
Deadline to present some
insights to the public
Open Belgium
Sharing our take
aways
2016 2017 April - Aug - Oct Nov Dec 2018 - Mar
Smart City Expo
Barcelona
Presenting the project
Team involved
Ramya
FARS data analysis, front
end
Bangalore
Bhargav Dana
Data analysis,
preparation
Bangalore
Testing additional
datasets
Washington, DC
Michele
Research, data prep,
statistics
PhD, King’s College
London
Ryan
Technical scope of the
project
San Francisco
Final team
Rasagy
UX design, data
visualization
Bangalore
Eric Morgan
Data modelling.
Creator of tippecanoe.
San Francisco
Data analysis, data prep.
Creator of turf.js
San Francisco
Mikel
Local coordination,
data prep
Washington, DC
Take aways
working with local
government on (open)
data-driven tools
MVPs
Prototype - “Show don’t just tell.”
Deadlines
Set external deadlines
Passion
Assemble a diverse team that’s passionate about the cause
Feedback
Give recommendations to improve data quality
Plan B
Have an alternative in case not all data is accessible
Newsletter mapbox.com/cities
Questions christina@mapbox.com

Traffic safety - answering tough questions with open data