Case study · Emergency alert system

Stadium emergencies, answered in seconds.

An emergency alert system for stadiums — a command console for the ops room and a responder app for staff in the field. One tap broadcasts across push, email and SMS, and live safety check-ins stream back so command knows who's safe.

Role  Lead Product Designer Domain  Emergency response Scope  Web + mobile · working prototypes
milvuseye — stadium alerts · command console

live render · desktop & mobile

01 · Problem

Spotting the emergency is fast. Telling everyone isn't.

A sold-out venue holds ~68,000 guests and roughly 300 staff — ops, contracted security, medics, ushers, concessions. When a fire warning or bomb threat hits, cameras and radios surface it in seconds. But turning that into one clear instruction reaching every person means juggling the PA desk, radio channels, a group email and ad-hoc texts — each with different wording, written under adrenaline. And once the message is out, command is blind: there's no way to know who received it, who's moving, or who needs help.

friction 01

Broadcast is slow and fragmented

Email lives in one tool, SMS in another, push in a third — if it exists at all. Someone composes instructions from scratch mid-crisis, and every channel ends up saying something slightly different. Minutes pass before the last staff member hears anything.

friction 02

Nobody knows who's safe

After the alert goes out, the feedback loop is a radio roll-call and a paper headcount. A steward trapped in a smoke-filled corridor looks exactly like one who simply didn't answer. Command makes evacuation calls without the one number that matters.

A packed stadium bowl seen from the upper stands
One venue, tens of thousands of people — and one instruction that has to land everywhere at once.
Two security cameras mounted on a concrete wall
Cameras spot the emergency in seconds — getting 300 staff to act on it takes far longer.

02 · Research

What shaped the design.

I studied NCS4 venue-safety guidance and public emergency-communication patterns (wireless emergency alerts, campus alert systems), then mapped the full alert-to-all-clear flow — from the moment command decides to act, to the last safety check-in. Four findings kept repeating, and each one became a design decision you can see in the prototype below.

No composing mid-crisis

Pre-built alert types, ready to send

Fire, bomb threat, medical, earthquake, active shooter, evacuation — each with prewritten, plain-language instructions. Under adrenaline you pick and send; you never write copy. Title, message and severity are filled in before the emergency happens.

Redundant by default

Email, push and SMS in one send

People miss channels, not messages. One broadcast fans out to every enabled channel with identical wording, so a steward in a loud concourse gets the SMS even when the push notification drowns.

Close the loop

Every alert asks: are you safe?

Recipients answer Safe or Need help with one tap. Check-ins stream onto a live board — 153 of 300 in, 31 need help — replacing the radio roll-call with a number command can act on.

Practice without panic

Live, Test and Drill are unmistakable

Venues rehearse constantly, and a drill mistaken for a real alert is its own emergency. Mode is a first-class, color-coded choice — Live is red and looks it — so no one ever wonders which world they're in.

A hand sketching a user flow diagram on a whiteboard
Mapping the alert-to-all-clear flow — every hop between decision and delivery is time lost.
A team working through notes pinned to a board
Synthesis: clustering emergencies into eight alert types, each with prewritten instructions and a channel plan.

03 · Design system

One language, two surfaces.

The console and the responder app run on one shared token set, so "critical" is the same red in the ops room as it is in a steward's hand. Severity is never carried by colour alone — every level pairs a colour with an icon and plain-language copy, so it survives a sunlit concourse, a cheap screen, and colour-blind eyes.

Semantic colour

Critical--red · #E74C3C
Warning · Need help--coral · #F39C12
Informational--blue · #3B82F6
Safe · Active--accent · #1ABB9C
Command chrome--navy · #2A3F54
Resolved--green · #27AE60

Each token ships a paired -soft tint for backgrounds and a -line value for borders, so a severity reads at any weight without new colours.

Typography

Attention!!! — Fire WarningDM Sans 700 · 26/1.15
Evacuation Notice · Sector BDM Sans 500 · 16
Please evacuate to the nearest assembly point.DM Sans 400 · 13
153 checked inInter 600 · 15 · app
10 sec ago · 300 recipientsInter 400 · 10.5

DM Sans for the console's data-dense chrome, Inter for the app — both chosen for legible numerals under pressure.

Alert types — one icon per emergency, prewritten copy behind each

Fire
Bomb
Medical
Quake
Shooter
Evac
Storm
Fight

Selecting a type fills the title, body copy and severity — under adrenaline you pick, you never write.

Components

Send Alert Resolve Cancel
LiveTestDrill ActiveCritical
Email Push SMS

Live is the only red mode — a drill can never be mistaken for the real thing.

Safety check-in bar

Safe · 115 Need help · 31 No reply · 7

The one component command actually watches — proportional, glanceable, and readable from across the room.

04 · Prototype

Two surfaces, one shared incident.

These aren't pictures of a prototype — they're the actual working prototypes, rendered live on this page. The command console runs the ops room: broadcast an alert, watch safety check-ins stream back, track tasks and recipients. The responder app puts the same incident in a field worker's hand. Launch either one in a new tab and click around.

Desktop · command console

Launch desktop prototype ↗
milvuseye — stadium alerts · command console

Mobile · responder app

Launch mobile prototype ↗

Live embeds of the real prototype files — use the launch buttons to interact with them full-size.

05 · Result

From "something's wrong" to everyone knowing — in seconds.

Modeled against the fragmented baseline — separate tools, hand-written instructions, radio roll-calls — the system collapses the two clocks that decide outcomes: how fast the alert reaches everyone, and how fast command knows who's safe.

< 10sFrom decision to alert on every phone — push, email & SMS in one send
↓ 90%Time to compose & broadcast, via pre-built alert types
LiveSafety check-ins replace the radio roll-call — who's safe, who needs help, in real time

Figures are illustrative of the concept's intended impact, not measurements from a deployed system.

Have a gnarly, high-stakes interface?

Emergency systems, ops consoles, AI copilots — I make the right action obvious under pressure.

Scene photography via Pexels · prototypes are interactive concept builds · figures are illustrative.