Geospatial insight.
Informed action.

The brief: problem, people, proof

Our SCADpro team worked with Deloitte to develop SAGE. As co-lead I owned project management, UX research and design strategy, shaping wireframes with the design team from research to a functional MVP.

Context from the project’s state research; counts from our research and testing.

What these teams already have — and why it isn’t enough

Before drawing a single screen we took apart the platforms these agencies use today. Five public systems, each read for what it does well, where it fails a decision-maker, and the opening it leaves behind.

• ArcGISDeep analytical tooling and secure architecture — and it needs training and a budget before anyone gets an answer out of it.
• NASA EarthdataSatellite coverage of the entire planet, free to anyone, with nothing to say about a single county.
• UN-Water SDG-6 portalClean visualisation across many categories; no state or regional detail, and it does not refresh while you work.
• GeoPlatformGovernment-backed and transparent, with an interface non-technical users struggle to move through.
• WaterSenseAuthoritative standards buried under a hyperlink tree.

We also looked outside government — mobility and foot-traffic analysis, indoor airport routing, stadium wayfinding, the digital twin of Paris built for the 2024 Olympics — to see what geospatial looks like when it is built for ordinary people rather than analysts. Real-time tracking, plain visual language and services shaped around a single task all came back into our own design.

Narrow the scenario to deepen the design

We initially explored Texas and five government-user archetypes. Deloitte’s feedback pushed us to narrow the user groups while widening the research context across California, Iowa and Texas.

Secondary research covered existing platforms, data sources, policy and resource distribution. Six expert interviews then helped the team understand data integration, emergency coordination and decisions made with delayed or imperfect information.

Widening the research meant three states instead of one: California, Iowa and Texas — top agricultural exporters with different climates and different disasters, which let us cross-check what food and water scarcity looks like across the country.

That is the argument for the scenario. A fire in Ventura County is not a county-sized event: it moves through processing, transport, distribution and retail until it reaches a shelf. California produces 95% of America’s processing tomatoes and 30% of the world’s; Iowa tornadoes flatten the corn that feeds livestock; Texas hurricanes take cotton and cattle with them.

Six experts, and the instruction they all gave

Secondary research tells you what exists. It does not tell you what it feels like to decide. We wrote a structured guide, kept the questions open, ran think-aloud sessions, and themed every answer into usage, pain points and opportunity.

• A data scientist“We don’t know what data we’re missing; we maximise what we have with tooling.” Fusing imagery, road and mapping data at different resolutions is still the core problem — and end users have to be able to trust the algorithm.
• A national security specialistA twenty-four-hour processing turnaround blocks instant decisions, and AI has to be trained carefully before it is trusted with sensitive calls.
• A space leaderSatellites give a cheap, repeatable revisit to a site — water levels, soil moisture, methane. Sustainability officers now need that data to report against climate goals.
• A chief sustainability officer-in-residenceLIDAR located the largest debris after a disaster and redirected crews — “in the world of emergency management, you can never get back time.”
• A geospatial technology leaderFEMA coordinates through the Incident Command System; the hard part is balancing household-level data against privacy, and cutting through the sheer volume.
• Two emergency managersFEMA lifelines — water, energy, health — are assessed first. Priority follows population, so rural areas lose to cities, and local windshield surveys still fill the gaps.

Six conversations, one consistent instruction: fewer surfaces, faster answers, and a visible reason to believe them.

What we were designing against

A brief this domain-heavy arrives with rules attached. Some came from the client, some from the data itself, and some from the people who would have to act on whatever we made. Every one of them left a mark on the interface.

“We are drowning in data; the real need is for fewer, more intuitive applications that can provide immediate, actionable answers for responders.”

Geospatial technology leader · expert interview

That line is the brief behind the brief. Every constraint above points the same way: less surface, fewer steps, and a reason to believe what the screen says.

Paper to pixels, and the note that changed the story

We started on paper — structure first, then a screen at a time — and took critique between every fidelity step rather than at the end.

The second checkpoint with Deloitte was blunter, and more useful: we needed a narrative. Not a feature set — a realistic scenario with a beginning, a turn and a resolution. A California wildfire. A short-term responder and a long-term planner. Consequences that land on food and water.

That note is what made the next chapter possible. Once the work had to carry a story, the question stopped being “what can this tool show?” and became “what does this person need next?”

The pivot: the user is not a “map person”

An early interface organized the experience around maps, reports and datasets. A guided review with a Deloitte managing director exposed a more useful framing: the user’s job was to understand the situation quickly and choose an action, not navigate a geospatial tool.

123

1. Report dashboardOngoing and priority reports compete for attention.
2. Progress reportsProject tracking takes the prime space.
3. Recently contactedPeople, not the emergency, anchor the side rail.

1234

1. Live alert firstThe active emergency sits on top, one click from its event map.
2. Ask, don’t huntAn AI search replaces digging through datasets.
3. Suggested questionsRecommended prompts give a one-click path.
4. Ongoing and predicted eventsReal-time reports sit beside AI-forecast events.

Decision: lead with the situation and a question, not a wall of tools.

1234

1. Tool-first sidebarFeatures before the situation.
2. Layer filtersUsers configure data before they see the event.
3. Analysis chartsScenario charts need expert reading.
4. Whole-state mapNo event in focus.

1234

1. Live event dataTemperature, wind and fire direction at a glance.
2. Colour-coded zonesTier 1 directly affected, Tier 2 potential spread, Tier 3 predicted future impact.
3. Recommended promptsThe next useful question is one tap away.
4. Recommended actionsSuggested from historical and real-time data.

Decision: open on the event area and put the next action beside the map.

Four rounds of expert and user feedback

Three guided expert reviews with Deloitte were followed by a classroom round with eight testers. Each round shaped the next revision.

The team redesigned the interface repeatedly rather than treating the first visual direction as settled; the final design reflects what each review revealed.

Access → analyze → act

SAGE connects an AI-powered search entry point with real-time event views, evacuation planning, resource allocation and post-event reporting. The Woolsey Fire scenario links those capabilities into one decision sequence.

1. AccessEssential geospatial data in one place, reached by asking a question.
2. AnalyzeRisk zones and predictive models turned into decision-ready visuals.
3. ActReal-time alerts and recommended actions for a coordinated response.

123456

1. Livestock routesFarm animals get their own evacuation paths.
2. Resident routesResidents are split into groups to ease road congestion.
3. Plan in chatThe AI drafts an evacuation plan from a plain question.
4. Confidence and sourcesA confidence level and the datasets behind the plan.
5. Recommended actionsAlert the public; relay the plan to local farmers.
6. Ask SageFollow-up questions without leaving the map.

12345

1. Shelter occupancyLive capacity for each shelter on the map.
2. Truck logisticsTrucks and distance per shelter.
3. Required suppliesFood, water and the timeframe to deliver them.
4. High confidenceSources linked for every figure.
5. Coordinate with FEMAOne tap to bring in federal support.

One fire, two people, ten days

The whole system had to hold together as a single sequence, so we ran it as one: the Woolsey Fire in Ventura County, from the first alert to the resilience plan that outlives it.

1. The alertA wildfire erupts and spreads. Sarah opens the event, not a map.
2. Read the situationLive event data shows which areas and which resources are most at risk.
3. EvacuateShe asks for a plan for at-risk residents and livestock, and gets the datasets and a confidence level with it.
4. Look past the fireWhile crews work, she checks its future impact on food and water resources.
5. DeployAt the peak, transport routes carry food and water to shelters, with live capacity and required supplies.
6. ContainedTen days later, damage is assessed and the impact on food and water summarised.
7. Hand overThe final report goes to Matt, the county’s Chief Sustainability Officer — with every question Sarah asked along the way.
8. Plan for the next oneMatt works from the same system to strategise for food and water resilience.

The handoff is the point. The Resource Manager’s ten days become the Long-Term Planner’s starting document — one system, one evidence trail, two horizons.

The outcome

The project delivered a functional MVP, which Deloitte carried forward as the foundation for the next phase.

The next step is task-based testing with resource managers and planners using realistic data latency and failure scenarios, measuring time to action, first-click success and trust in confidence labels.

My takeaway: when the stakes are high, a richer map is not automatically a better interface. The design has to connect the evidence, its uncertainty and the next action.

Three directions were set out for a next phase: a platform that performs across devices so on-ground agents get the same edge; the same data democratised for farmers and communities rather than agencies alone; and immersive visualisation for the point where a flat map stops being enough.

Made by a multidisciplinary team

Teamwork made it possible to connect domain research, interface design, visual storytelling and a working prototype across 15 students and 7 majors.