CASE STUDY: Youth Without Shelter

CASE STUDY: Youth Without Shelter

“We have never stopped with our vision. We want to end youth homelessness one youth at a time and that is our focus. The people that come through our doors are individuals and we treat them as such and we want to help them overcome the struggles they are experiencing.”
– Steve Doherty, YWS Executive Director

Who is Youth Without Shelter?

Youth Without Shelter’s Mission statement is, ”Empowering youth facing homelessness to reach individual potential.”

Youth without Shelter (YWS) was founded in 1986 by a group of teachers and guidance counsellors who were frustrated with the lack of safe, emergency housing and support programs for their students. YWS has now helped more than 15,000 homeless youth to build their confidence and life-skills and to find long-term stable housing and jobs.

An impressive 87% of every dollar donated is directed to youth programs and related support (remaining 13% for admin and fundraising).

YWS has been nationally recognized for their work and most recently as Charity Intelligence Top Charity Pick and as a Money Sense 2019 Top 100 Charity.

Problems we were asked to solve.

There were four primary questions YWS asked us to tackle during the datathon.

  1. Generate a map of segmented donor profiles of current YWS donors using geographical data from StatsCan and existing donor data.
  2. Validate the hypothesis of whether volunteering is a gateway to donating by matching volunteer data to donor data and if there are any insights about these volunteers who do become donors. E.g. Do they donate more?
  3. Identify the most loyal and consistent givers in the past 5-10 years such as those who have given at least 7 times in the past 5 years; and those who have cumulative giving of $1,000 over the past 5 years to aid in major gifts program launch.
  4. Identifying the profiles of lapsed and new donors. For lapsed donors find a pattern of why they have lapsed and who would likely return. For new  donors find a pattern of what channel they first became a donor and any patterns for repeat donations.

What did the dataset look like?

YWS provided an extensive set of donation and volunteer data spanning 20 dimension and close to 100,000 rows of data. The earliest records date back to 1989!

After cleaning and preparing the data from YWS, the resulting tables including the following data types:

  • Unique Constituent ID
  • Key Indicator
  • Gender
  • First Gift Amount
  • First Gift Date
  • Last Gift Amount
  • Last Gift Date
  • Largest Gift Amount
  • Largest Gift Date
  • Total Gift Period (days)
  • Donor FSA

  • Donation_approach
  • donation_designation
  • gf_type
  • volunteer_hours
  • volunteer_events
  • is_volunteer
  • is_constituent
  • Is_donor
  • Days_between_donations
  • was_Volunteer
  • Volunteer_TotalHours
  • Volunteer_MostRecentDate

Next, StatsCan Forward Sorting Area tables were used to associate postal code information with their geographic areas in Canada. This allowed us to readily add socio-economic and demographic information for each FSA. The data included details such as total population, number of people in each income bracket, age breakdowns, etc…

To assist Data For Good volunteers, several data dictionaries were created to provide metadata about each dimension of the data provided by YWS for analysis.

“There are 2000 youth experiencing homeless every night in the city of Toronto and less than 400 shelter beds and transitional housing beds. It is a challenge we continue to experience and it is going to take imagination and people like yourselves to help make a difference so thank you for making a difference.”
– Steve Doherty, YWS Executive Director

The tools and analysis we used to help

Multiple tools were used, including Tableau, R, Python / Jupyter Notebooks, PowerBI and Excel.

Several different approaches were used to analyze the data. Each group of Data For Good volunteers chose their own approaches. Some of the models put into use are listed here:

  1. Logistic Regression
  2. Lasso & Ridge Regression
  3. Random Forest
  4. Support Vector Machines
  5. Naïve Bayes

Key Findings and Recommendations

Some of the most interesting findings came from analyzing donor profiles and creating segments to better understand who YWS’ most loyal donors are.

Donor Distribution by Gender

Donor vs Volunteer Location

Total Donations by FSA Code

Recommendations included focusing fundraising efforts around donors with the following attributes:

  • Female groups in urban, middle-income areas who are recent joiners
  • Potential campaign days: Mother’s Day, International Children’s Day, Donor’s Birthday

Other significant findings looked at volunteering and how it correlates with donations.

  1. 5% of Constituents are Volunteers and 4% of them are Donors
  2. Volunteers donated 28% of the overall amount
  3. Organizations tend to volunteer and donate more than individuals
  4. Volunteers and donors come from the GTA

Volunteers Donate more, and Donors also Volunteer more

 

 

Acknowledgements

These events are made possible by the support of amazing sponsors. Data For Good and YWS are very grateful to Capco who generously volunteered the use of their space, resources and most importantly food!!

Case Study questions? Ask stephen.boyd@dataforgood.ca

Interested in working with Data For Good? Get in touch with us! contact@dataforgood.ca

 

2019-06-04T22:34:01+00:00