Understanding SF Homelessness Through Data and Human Stories
Walking in San Francisco
A research project driven by curiosity and carried out with passion, pairing public data with the human stories behind it.
In San Francisco, losing a job can mean losing housing within weeks. Generated with Gemini Nano Banana Pro.
Why this problem matters
Timing matters more than income. The risk of homelessness is less about how much someone earns and more about how fast support reaches them after a shock.
Who is affected
Big tech employees and manufacturing labor alike. Our target user group was people who lost their job within the last 30 days, the window where a small intervention can change the outcome.
What prevents people from accessing housing support fast enough to avoid homelessness?
Research and methods
We went wide before narrowing: people on the street and in shelters, the orgs that run the response, and the public budget and policy underneath it.
Interviews, more than 50 people
We met people on the street, in the shelter, and who had experienced homelessness. More than 50 in total.
Shadowing and org interviews
To understand the higher level, we contacted the Coalition on Homelessness, the Homeless Outreach Team, and people working for the government, and met them in person.
Secondary research, budget and policy
To learn how the system works, I dug into the public documents and came to understand the pros and cons of Coordinated Entry.
Key insights
The system reacts too late. Housing support should intervene during instability, not after collapse.
- 1.Time is the real bottleneck
- 2.Access is not the same as awareness
- 3.Rules punish instability
- 4.Shelters are not perceived as a safe fallback
How might we help newly unemployed SF residents access temporary housing support within the first 30 days, before they fall into long-term homelessness?
Solutions
Designing leverage points to change system behavior. The timeline runs job loss, then housing loss, then homelessness, so the intervention is a 30-day housing bridge placed before housing loss becomes permanent.
Access by time risk, not paperwork
Impact
Access is granted based on time risk rather than eligibility documentation.
Why this matters
In existing systems, delays often push people past the point where recovery is possible. By intervening early, the system changes when housing support begins, not just who receives it.
Temporary housing, framed as transitional
Impact
Users gain immediate stability without losing clarity about what comes next, reducing both dependency and disengagement.
Why this matters
Research showed that unclear shelter rules often discourage participation. Making the temporary nature explicit builds trust while preserving forward momentum.
A deliberate re-entry into formal systems
Impact
Users return to caseworkers, documentation, and coordinated entry with clarity and agency rather than exhaustion.
Why this matters
Systems designed for stability often fail people during instability. This design creates a deliberate re-entry point.
Reflection
Leverage to change the system, not solve the problem
This project taught me that effective system design is not about solving the entire problem, but about finding the right leverage to change how the system behaves. Rather than attempting to “fix homelessness,” I focused on a specific intervention point, the first 30 days after job loss, where a small shift in timing and access could significantly alter long-term outcomes.
Designing dignity into systems
Through interviews, I learned that many people avoid shelters not because they reject help, but because current systems often require them to give up autonomy, belongings, or clarity about their future. This insight pushed me to treat dignity as a functional requirement, not an abstract value.