Brickbybrick's SilverWatch

Inspiration

Our inspiration for SilverWatch comes from our own experiences volunteering with the Agency for Integrated Care (AIC), where we had the opportunity to interact directly with elderly residents in the community.

These experiences made us more aware of how quickly Singapore's population is ageing and how some seniors may face challenges such as living alone, limited mobility, reduced social interaction, or difficulty accessing community support. What stood out to us was that vulnerability is not always immediately visible. A senior may appear independent while still lacking regular social contact or a nearby support network.

At the same time, we saw firsthand how valuable community outreach can be — but manpower, time, and community resources are limited. This led us to a simple question:

If we only have 20 volunteers this Saturday, where should they go to make the greatest impact?

SilverWatch was born from the idea that data can help community organisations answer this question more systematically and ensure that limited volunteer capacity is directed towards areas where it is most needed.

What it does

SilverWatch is a decision-support tool for planning elderly community outreach.

Rather than simply showing where elderly residents live, SilverWatch helps coordinators translate community data into an actionable deployment plan.

It combines indicators of elderly need and access to community support into a transparent Priority Index, highlighting neighbourhoods where outreach may be most valuable.

SilverWatch then uses an optimisation model to recommend how available volunteer teams should be allocated while respecting real-world constraints such as manpower, travel time, permitted service areas, and team availability.

The platform follows four simple steps:

Identify → Optimise → Explain → Simulate

Identify neighbourhoods with higher potential outreach needs.
Optimise volunteer deployment based on available resources and operational constraints.
Explain why areas are prioritised, skipped, or limited.
Simulate how deployment plans change when volunteer availability or constraints change.

Ultimately, SilverWatch aims to answer one practical question:

Where should we send help first?

How we built it

SilverWatch is designed as an end-to-end solution on Databricks.

We combine Singapore open datasets covering elderly population characteristics, living arrangements, community support locations, and geographical boundaries.

The data flows through a Bronze → Silver → Gold architecture:

Bronze: Raw public datasets are ingested from sources such as data.gov.sg, SingStat, HDB, MSF/AIC, and OneMap.
Silver: Data is cleaned, standardised, geocoded, and joined across geographical areas.
Gold: Analysis-ready priority and coverage features are generated for optimisation and visualisation.

We then construct a Priority Index using indicators such as elderly population concentration and accessibility to community support.

A Mixed-Integer Linear Programming (MILP) optimisation model using SciPy/HiGHS allocates whole volunteer teams across eligible neighbourhoods while respecting operational constraints.

The results are presented through an interactive interface where coordinators can view priority areas, recommended deployments, explanations for allocation decisions, and test different volunteer-availability scenarios.

Challenges we ran into

One of our biggest challenges was translating a complex social problem into something that could be modelled responsibly.

Vulnerability cannot be perfectly captured by a single number. Public datasets provide useful neighbourhood-level indicators, but they cannot tell us whether an individual senior is vulnerable or requires assistance. We therefore designed SilverWatch as a decision-support system, rather than a system that automatically labels individuals or replaces the judgement of outreach coordinators.

Another challenge was combining datasets collected at different geographical levels and ensuring that our indicators could be meaningfully compared.

Finally, optimisation required us to balance two competing goals: directing volunteers towards areas with greater potential need while still respecting practical constraints such as team availability, permitted areas, and travel time.

Accomplishments that we're proud of

We are proud that SilverWatch goes beyond creating another heatmap or analytics dashboard.

A heatmap can tell us where potential need is. SilverWatch helps decide what to do next.

By connecting community data directly to volunteer allocation, SilverWatch turns insights into an actionable outreach plan.

We are also proud of keeping the system transparent. Coordinators can understand why an area was prioritised, why resources could not be allocated elsewhere, and how the recommendation changes when assumptions or volunteer availability change.

Most importantly, the project combines something personally meaningful to us — our volunteering experiences — with our background in data analytics to address a real operational challenge in community care.

What we learned

Building SilverWatch taught us that applying data analytics to social problems requires more than simply building an accurate model.

Data needs context. Optimisation needs constraints. And recommendations need human judgement.

We learned the importance of designing analytics around the decisions that users actually need to make. Instead of beginning with "What model can we build?", we began with "What decision does an outreach coordinator need help making?"

We also learned that transparency is especially important when working with social-impact applications. SilverWatch should help coordinators understand and explore possible deployment strategies, not hide decisions behind an algorithm.

What's next for Brickbybrick's SilverWatch

Our next step is to develop SilverWatch from a proof of concept into a working pilot.

We plan to expand the data pipeline across more neighbourhoods, refine the Priority Index with additional community indicators, and conduct sensitivity testing to understand how different weighting assumptions affect recommendations.

We also aim to develop the What-If Simulator, allowing coordinators to ask questions such as:

What if 10 more volunteers are available?
What if volunteers can be shared across service areas?
Which neighbourhoods would still have unmet outreach needs?

Finally, we hope to validate SilverWatch with real outreach workflows and measure whether assisted planning can reduce planning time while increasing the amount of priority-weighted community need reached with the same volunteer capacity.

Our long-term goal is simple:

Same volunteers. Smarter deployment. More need reached.