One of my favorite things about our work in data analytics has been the chance to support varied parts of the City do more with their data. In 2018 we partnered with the Seattle Municipal Court (“SMC”) to explore how we could improve the outcomes of parking and traffic tickets issued in the city.

This is no small issue: SMC issues some 600,000 parking tickets and traffic camera citations per year. Joint analysis of 2016 data showed that some 40% of issued tickets defaulted, and some 25% of tickets were referred to debt collection agencies; bad outcomes for drivers and for SMC. We wondered whether a redesign of communications – informed by behavioral science, similar to our work on pet licensing – could help, and whether we could leverage this same administrative data to track a low-cost but and robust randomized experiment.

Our approach

Motivated by a desire to reduce default and referrals to debt collection, we designed two new communications:

1. a new version of the default reminder card sent to each driver after 19 days; and
2. a new courtesy notice reminding drivers of their payment obligations that would be sent after 9 days (prior to any default).

Old (left) and redesigned (right) default reminder card

New courtesy notice

We ran two simultaneous randomized control trials (“RCTs”) to evaluate the impact of these changes. This approach would allow us to test not only the individual impact of our two interventions was incremental.

The results

The results of the trials, conducted over 10 weeks, showed significant impacts on ticket outcomes. Even controlling for impact of the other intervention, the new courtesy notice demonstrated a 13% reduction in the likelihood of tickets defaulting and the new default notice design demonstrated a 9% reduction in the likelihood of tickets ending up in debt collection.

These outcomes are more than just interesting statistics. The new designs would lead to about 22,000 drivers annually – including around 8,000 people of color each year – to avoid debt collection over an unpaid ticket.

Next steps

Inspired by some fantastic examples in other cities, we’re looking to deepen our use of behavioral science, data and design to improve City services. We’d love to hear of ideas and examples of where some of these approaches could be used to better-support those who live and work in Seattle.

Guest post by Namrata Kolla – a proud IP alumnum.

Nam finds herself well-positioned under wall-mounted arrow

On the first day of my internship, my manager took me to Caffe Vita[1], a coffee shop a few blocks from City Hall. When I asked him which projects I would have the chance to work on, he slapped a white sheet of paper on our tiny tabletop and said, “Here’s a Kanban!”.

My first thought was, “What’s a Kanban?”

My second thought, “Holy smokes, that’s a lot of options!”

Until I started my internship, I did not realize the breadth of services that municipal governments provide. I had some vague ideas via my interactions with the City, but through my Data Analytics internship I suddenly saw many more. Below are three takeaways from my time as an official Seattle government insider:

1) City government collects a lot of awesome, important data.

Just last month, I was listening to a presentation where my professor stated that the three organizations that have the most data in the world are Google, Facebook, and Amazon. I spoke up in class that day to tell the professor he forgot a big one: government!

Our city has some of the richest datasets in the world. While there are valid reasons to be wary about data collection and storage, government data, in particular, are pretty special. My favorite reason is that government data tend to be more inclusive than what most private companies can generate. For example, one of Amazon’s biggest datasets is the purchasing behavior of millions of people. At the same time, in the service of making its services better, Seattle City Light collects data about the quality of its power supply to almost every household in Seattle – not just those people that have computers or use particular websites. To protect this data, the City has some of the strongest privacy and data security policies in the world.

The power and extent of administrative data was a big lesson during my internship. There are many obstacles in the way of using administrative data to their full potential, some of which were placed with purpose. But, I feel that too few administrators are aware of how much more certainty and nuance they could have in their program decisions if they better (and safely) employed the data we already have at hand.

2) Analytical innovation will never be as important as trust and communication. 

After spending 70% of my time crunching numbers on a computer screen, 20% percent talking to partners in meetings (and 10% getting coffee), I found time and time again that the conversations I had in the 20% of my time decided the fate of my projects far more than any discovery I made in that 70%.

While getting desk time with no distractions is definitely something to relish, analytical problems are always easier to solve than organizational problems. When I ran into a bug trying to manipulate shapefiles from the American Community Survey, I spent an hour brainstorming with a colleague in public utilities, two hours on the phone with the University of Washington’s statistics department, and at least four hours with Dr. Google. When I ran into hurdles trying to access data to complete a project for a department partner, I spent an hour sending emails and three hours in discussions. While the first problem may have been more laborious, I never felt as stressed as I when I was trying to solve the second problem. I usually feel confident that I can solve a technical problem or find someone to help me, but I am rarely confident about problems requiring motivating people. Fortunately, I have been able to observe my colleagues in the Innovation & Performance (IP) Team and the City Budget Office and learn more about how to communicate intent and value building relationships over short-term success.

 3) Analyzing data in government doesn’t have to be a lonely job.

I heard myself and several City data analysts mention feeling lonely at some point or another while trying to solve day-to-day problems. While this is not unique to our role, often there are only a handful of people who use data in the particular way that we do and there is nobody else to ask for help. The silos that naturally grow from the organizational structure of the City also made it hard for me to find help in other departments when I initially started my internship.

However, I was lucky that the IP team organizes monthly City Performance and Data Science meetings where I had a chance to meet city workers in similar problem spaces. Additionally, the nature of our work being conducted in departmental partnerships helped me build authentic connections with other analysts while working in the weeds. Nothing cultivates a relationship with another person better than going through something hard, and I have found that to be as true in city government as in my personal life!

Looking back at my time on the IP team, I am so grateful to have been a part of this family. It has made a huge difference in my life to see city workers so passionate and committed to bettering communities and making long-term, systemic change. While there were countless challenges, there were many more celebrations and I have only become more humbled and excited to pursue a career in public service.

[1] Other excellent coffeeshops are available and are indeed patronized by the IP team

Last month, Tina and I attended the first Behavioral Cities Summit organized by the Kahneman-Treisman Center for Behavioral Science & Public Policy at Princeton University. The Center hosted representatives of five cities committed to the use of behavioral science to improve city government.

No shiny presentations to share

The event was different from many conferences in that there were few project case studies and little self-congratulation. Rather, the Summit was built around frank discussion of what had and hadn’t worked for each team: from day-to-day operating models to managing external partnerships. Our shared hope was that the connections made during that day would form the basis of a network of practitioners who could share learning across our different cities.

I was most interested in learning how other cities had scaled their work. New York and Chicago have embedded teams from Ideas42, a leading non-profit design and consulting firm in the field. GovLabPHL has pursued a strategy of partnership with local academic institutions, bringing the insights of expert researchers to the way that that the City of Philadelphia communicates with its residents. In a time of significant budget pressure for the City of Seattle, we believe that the use of behavioral science could improve the effectiveness and efficiency of the City’s work across many domains, and it was inspiring to see other cities deploying this expertise to such ends.

Photocredit: Anjali Chainani, City of Philadelphia

Expanding the amount of technical expertise available to a city government is only half the story. The inspiration for each of our projects came from our colleagues within the City of Seattle; people who think daily about how to improve the way that the City works and communicates. Our job is to bring those ideas to life. Next year, I want to make sure that we reach more teams in more departments to explore how we could help them achieve their goals.

Looking forward

So what will be different? As we plan for next year, we are going to explore options for scaling our work; bringing more external expertise to the City and engaging more of our colleagues. We will make it easier to engage with our team and share ideas. As we develop these approaches, we will share them on this blog. Let us know if you have any ideas!

What is it?

Earlier this year, our team reorganized its work around three service offerings for the City. A central theme within each was the desire to bring more and better data into day-to-day decision-making.

Shortly afterwards, we met with the Seattle Fire Department (“SFD” or the “Department”). Led by Chief Scoggins, the Department is a national leader in a range of efforts to improve services and better serve the people of Seattle. In this vein, SFD had identified an opportunity to drive performance with own operational data, but faced a series of obstacles:

• Disparate and inaccessible data sets. Operational data was stored in thirteen unconnected data systems, in a wide variety of formats.
• Reporting tools that for worked for some, but not all. Many of SFD’s systems for monitoring and interrogating performance were static spreadsheets containing Department-level data. Many found them difficult to interpret; those who had narrower operational responsibilities often could not easily isolate the performance of their battalion or unit.
• Few opportunities to bring data to operational problem-solving. With few regular opportunities for Department-wide interrogation of operational measures, performance management often relied upon ad-hoc communications from leadership to operational groups.

Our Approach

In response, SFD and IP worked together to unlock the value in the Department’s existing data. Inspired by examples nationally and locally, we worked with SFD to design and deliver a STAT-style performance program for the Department, built around three efforts:

• Infrastructure. In partnership with Seattle IT, we automated and transformed key datasets among the disparate data sources maintained by the Department, providing a platform for real-time access and future analysis.
• Interactive performance tools. Informed by interviews with command and operational staff across SFD, we created operational dashboards that share and allow the interrogation of key operational and performance data over time. To meet the needs of operational staff and from department to battalion-, station-, unit- and even shift-level.
• Drawing upon principles of human-centered design, we have designed, set-up and prototyped a departmental STAT meeting for SFD. Designed with the unique shift pattern of a fire department in mind, we continue to iterate with the benefit of user experience.

Example of the interactive dashboards built with Seattle IT and the SFD

The Results

The first, prototype FireSTAT sessions have taken place in the last months and we continue to refine and develop this practice. Already, the project is bearing fruit for the Department:

• Deepening use of data and performance measures in by SFD management. The back-end automation and visualization of data has allowed existing performance measures to be better-integrated into operational management.
• Enabling SFD operational decision-makers to identify, analyze and respond to key trends in fire and EMS demand. Real-time plotting and evaluation of department operations has allowed emerging trends to be tracked, analyzed and responded to within days. Recent trends in illegal burning and hotspots for EMS call volumes were analyzed with FireSTAT.
• Supporting collaborative problem solving around key SFD KPIs. The new dashboards have allowed surfacing of relevant trends and comparison in nationally-recognized performance measures; prototype STAT meetings have facilitated shared problem-solving around these issues.

The first prototype FireSTAT meeting. Out of tiny acorns…

All this is only the beginning. We are excited to deepen the practice of FireSTAT and to expand its reach into other operational datasets. Watch this space!