DataScienceSF

What is DataScienceSF?

DataScienceSF is a way for you to harness the power of advanced analytics and applied statistics to implement a tool that helps improve your work. 

This service from DataSF aims to help departments achieve more with their existing resources and processes.

Through a 4 month engagement, DataSF’s Data Science team and your department will refine a problem, identify statistical methods to address it, and develop and implement a service change tool to improve your work. Projects are chosen through a bi-annual selection process. The final product of DataScienceSF isn't a recommendation or a report but a change to how services are delivered.

 

What types of tools does DataScienceSF use?

DataScienceSF will bring 3 key tool sets to bear on your service change questions:

  • Statistical Methods. DataScienceSF will help identify the right method for your problem. Methods we may use include sentiment analysis, machine learning, regression, data mining, classification, clustering, imputation, AB testing, forecasting and more.
  • Tools. We use a variety of languages, libraries, data engineering and visualization tools. Languages include python, R, Javascript, NodeJS and SQL with a variety of libraries (SciPy, Pandas, etc). Our data engineering tools include profiling, ETL, noticing, APIs and optimized data pipelines and storage/access. Visualization tools include D3.js, Gephi, Leaflet, PowerBI, ggplot2 and Shiny.
  • User Experience Research. Key to a successful data science project is the implementation of a service change via a data driven tool. We use a variety of tools to assess and help design the right implementation including journey and process mapping, service blueprinting, ethnographic research and ride-alongs, iterative prototyping and usability testing.

Collectively, these tools allow us to create an actionable data insight to improve your work.

What types of problems can DataScienceSF solve?

There are five basic types of civic problems data science can help address. Ask yourself if you or people in your department would want to:

Image showing target categories, people and places

Find the needle in the haystack

Do you have trouble identifying targets in a larger population? If you find it difficult to identify people, geographic areas, or categories to target, data science can help.

Image showing a backlog being cleared

Prioritize your backlog

Do you have a backlog that you tackle via first in, first out and are worried that you’re missing priority cases? Data science can help identify high, medium, and low priority cases by analysing existing data.

Image showing someone heading to a bad outcome

Flag "stuff" early

Do you find it hard to predict future conditions leading to reactive services? Many situations - good and bad - could be addressed more efficiently if caught early, even before they come to you. Data science identifies candidates for early intervention and engagement.

Image showing 2 separate forms

A/B test something

Do you use costly outreach methods (like mailers, text messages, forms, surveys) that aren’t tested before implementation? Data science can help identify and test various approaches to identify those that would be most successful.

Image showing an alternative distribution of resources

Optimize your resources

Do you find it difficult to identify where to place or distribute resources to be more effective? Data science uses existing data to optimize distribution of services (people, resources, equipment) to minimize response time and maximize throughput.

You can learn more about the different types and see examples of civic data science projects from other cities in our PowerPoint deck.

What are the expectations for department partners?

Below are some general expectations we have for the project. We will refine and clarify roles and responsibilities via a project charter.

  • Commitment to service change. First and foremost we expect departments to be open to and commit to a service change if that's where the data leads us.
  • Department Champion. To be an effective partner, we will need a Department Champion. The Department Champion is our liaison within the department who helps refine the problem statement and identify paths to service change. We estimate the Departmental Champion will need to dedicate on average 5 hours a week to this project. This will probably taper off after the beginning.
  • Commit to working within our timeframe. Engagements last between 1-4 months. The engagement will be an iterative process as we refine the original problem statement into something answerable with the data available and implementable within any service change constraints. If now is not a good time, we will have future cohorts.
  • Access to staff and business processes. As part of the engagement we expect to do user research. This may include 'ride-alongs', where members of the data science team shadow employees, interviews or other research methods. This will help identify relevant implementation factors that should be incorporated into the statistical model.
  • Timely data access. Any data science project relies on timely data access. Please plan on preparing and working with your technology or database team to provide data. As appropriate, we can sign data sharing MOUs (we have templates). Some projects may require ongoing data access to implement the model in real time.
  • Presenting and Disseminating. Part of our goal is to document and communicate what we learn and accomplish so both other departments and jurisdictions can benefit. We will need input and feedback and some participation from departments in this process. 

Application Process

 

Next Steps

Applications for Cohort 5 (FY 2022-23) are now closed. Finalists will be announced in November. If you want to find out more about the program and the next cohort, please email [email protected]

Cohort 5 applications are now closed

Applications for DataScienceSF Cohort 5 (FY 2022-23) are now closed. Finalists will be announced in November 2022

 

Resources

One-pager overview of DataScienceSF (PDF)

DataScienceSF Deck (PDF), including:

  • What is (and isn't) data science
  • Detailed info on problem types with real world examples
  • Overview of phases
  • Selection criteria

DataScienceSF Showcase of Past Projects

 

Other Resources: MOCI's Civic Bridge

  • Civic Bridge recruits pro bono private sector talent to work alongside government employees on critical City issues. 
  • Pro bono teams can offer services in user research and design, communications, strategic planning, facilitation, and more.