High-Throughput Checkout for Stadium Environments

A High-Throughput Checkout System for Peak Crowds

High-Throughput Checkout for Stadium Environments

My Role

My Role

Lead Designer 
 Feature Scoping, Research, Interaction Design, Prototyping, UI Design, Lottie Design, Dev Handoff

Lead Designer 
 Feature Scoping, Research, Interaction Design, Prototyping, UI Design, Lottie Design, Dev Handoff

Team

Team

Lead Designer 
3 Backend Engineers
2 Frontend Engineers
3 Product Managers
2 Devops
3 Data Engineers

Lead Designer 
3 Backend Engineers
2 Frontend Engineers
3 Product Managers
2 Devops
3 Data Engineers

Duration

Duration

4 Months

4 Months

Overview

Overview

Stadium checkout operates under extreme variability and compressed demand.

Transactions surge in short, high-pressure windows during pre-game and halftime, while lighting shifts from daylight to floodlights within the same event.

Multiple games in a single day introduce rotating menus, changing inventory, updated pricing, and event-specific branding in real time.

At the same time, refunds, receipt requests, and staff interventions must occur without slowing throughput, and customers are often distracted and in a hurry. 

Checkout at RadiusAI evolved in three phases. The first two phases focused on convenience stores across North America, piloting with partners like QwikTrip, CVS, and Catalina to stabilize checkout for packed and unpacked items at stores.

Phase 3 shifts to stadium environments, where transactions are high-frequency, queues are long, and cashier dependency must be minimized. The system must handle refunds, receipts, varying lighting conditions, and venue-specific menus and branding—all while serving crowds of up to 50,000 people within limited time windows.

HARDWARE CONTEXT

HARDWARE CONTEXT

Interaction Architecture in Stadium Environments

Interaction Architecture in Stadium Environments

Stadium checkout operates on a camera-powered, dual-screen system built for speed and operational resilience.

Checkout at RadiusAI evolved in three phases. The first two phases focused on convenience stores across North America, piloting with partners like QwikTrip, CVS, and Catalina to stabilize checkout for packed and unpacked items at stores.

Phase 3 shifts to stadium environments, where transactions are high-frequency, queues are long, and cashier dependency must be minimized. The system must handle refunds, receipts, varying lighting conditions, and venue-specific menus and branding—all while serving crowds of up to 50,000 people within limited time windows.

Customer Screen

Customer Screen

Ensures transparency items, pricing, tips, and payment.

Ensures transparency items, pricing, tips, and payment.

Operator Screen

Operator Screen

Provides absolute control to cashier or admin for inventory and over-rides.

Provides absolute control to cashier or admin for inventory and over-rides.

Camera Layer

Camera Layer

This layer continuously captures and interprets items placed in the checkout zone,
enabling real-time detection and pricing.

This layer continuously captures and interprets items placed in the checkout zone, enabling real-time detection and pricing.

This layer continuously captures and interprets items placed in the checkout zone, enabling real-time detection and pricing.

UPDATING THE FLOW

Starting with the backbone of project

Picking up the learnings of Retail POS

Below is a prototype for the retail POS, originally designed for convenience store environments and aligned with the typical retail customer mental model. The PM team has since adapted these fundamentals to support high-speed stadium checkout scenarios.

NO SUPPORT ITEMS

Incomplete cashier control limits awareness of customer stage and checkout progress.

Incomplete cashier control limits awareness of customer stage and checkout progress.

Checkout not customer-driven increases load and places full operational burden on cashiers.

Checkout not customer-driven increases load and places full operational burden on cashiers.

Dynamic menu inventory constantly shifts requiring real-time adaptability.

Dynamic menu inventory constantly shifts requiring real-time adaptability.

Last-minute product additions disrupt flow and introduce processing delays.

Last-minute product additions disrupt flow and introduce processing delays.

SLOW RESPONSE ITEMS

Unknown item handling delays processing and prevents system learning.

Real-time product remapping must occur instantly when misclassifications happen.

Frequent branding and layout shifts must adapt seamlessly without optional delays.

DISCOVERY

Before defining the problem, our Head of Sports & Entertainment (VP) shared live stadium insights on where checkout slowed down.
These revealed structural constraints that outlined the problem statement

UNDERSTANDING CHECKOUT BEHAVIOR

Major impact on stadium checkout speed

Stadium checkout delays are largely behavioral, not technical.

Retail POS insights show that queue distractions, order changes, and checkout questions drive most transaction interruptions.

30–35%

30–35%

Customer distraction in line

Talking, watching the game, or using phones instead of preparing for checkout.

20–25%

20–25%

Finalised order changes

Adding, removing, or swapping
items at the counter.

15–20%

15–20%

Questions during checkout

“Where is this?”,
“Can I get one more?”,
“Is this available?”

5–10%

5–10%

Unknown or misclassified items

Especially during peak selling moments.

3%

3%

Refunds or receipt requests

Interrupting the active checkout flow.

Design Observations, Finally!

30–35%

30–35%

Customer distraction in line

Talking, watching the game, or using phones instead of preparing for checkout.

UX Decision Taken

UX Decision Taken

Reduce dependency on customer-led interactions

Customer-led interactions

Select tip (Customer selects)

Tap “Pay Now” (Operator can initiate)

Tap or swipe card (Operator can initiate)

Select or identify unknown items (Operator can assist)

Design Implementation

Select tip (Customer selects)

Seamless "Tip" Selection During Checkout

Tip is pre-selected by default to reduce interaction and speed up checkout.
Customers can modify it anytime using the Custom Tip option.

Tip is pre-selected by default to reduce interaction and speed up checkout. Customers can modify it anytime using the Custom Tip option.

Customer Screen

Tap “Pay Now” (Operator can initiate)

Cashier's can tap "Pay Now" for the users

Cashiers can take over the checkout if customers are distracted.
This allows them to seamlessly guide the transaction through to payment.

Cashier Screen

Customer Screen

Select or identify unknown items (Operator can assist)

Cashiers can replace misclassified or unknown items using the quick menu.

Quick menu replacement avoids switching to barcode scanning.
The item can be replaced instantly during checkout.

Cashier Screen

20–25%

20–25%

Last-minute order changes

Adding, removing,
or swapping
items at the counter.

15–20%

15–20%

15–20%

Questions during checkout

“Where is this?”,
“Can I get one more?”,
“Is this available?”

UX Decision Taken

UX Decision Taken

Cashiers can answer customer questions through menu lookup and modify orders by adding or removing items.

Design Implementation

Cashier can manually "add" or "delete" item

Cashiers can browse the full menu to manually add or remove items that are physically present but not detected by the scanner, ensuring checkout continues without stalling.

Cashier Screen

5–10%

5–10%

Unknown or misclassified items

Especially during peak selling moments.

UX Decision Taken

UX Decision Taken

No fallback to barcode, instant correction

Instantly resolve by assigning this to an existing menu item.

Design Implementation

No fallback to barcode, instant correction

Cashiers can browse the menu and assign the correct item to resolve unknown scans. If the issue repeats, the cashier can train the item for future recognition.

Cashier adding item

Cashier training item

3%

3%

3%

Refunds or receipt requests

Interrupting the active checkout flow.

UX Decision Taken

UX Decision Taken

UX Decision Taken

Streamlined void/refund flow with limited-step confirmation

Receipts can be shared on demand by the cashier.
Customers can also collect them via the RadiusAI entry point shown for 5 seconds after payment.

Receipts can be shared on demand by the cashier. Customers can also collect them via the RadiusAI entry point shown for 5 seconds after payment.

Design Implementation

Cashier can manually "print receipts" or "refund" transactions

Cashiers can browse the menu and assign the correct item to resolve unknown scans.
The system learns from this correction for future recognition.

Cashier sorting "transactions"

QR code for Customer to collect "Receipt"

GAME ANALYTICS

At State Farm Stadium, a single game translated into over 7,700 transactions and nearly $55K in revenue — not bad at all for a first pilot event, especially one fueled by hungry fans and halftime rushes.

MOMENT OF MOMENTUM

The Business Saw an Opportunity!

The early pilot results didn’t just prove the system worked — they became the pitch to other stadiums.

Within the same season, three additional stadium partners signed on for pilot deployments, including Petco Park and the San Diego Padres.

LITTLE WALL OF FAME

MAJOR CHALLENGE

Fast turnarounds, shifting game-day conditions, and branding that had to show up instantly — with multiple games in a day, the experience had to keep pace with changing menus, and venue identity.

Solving Branding for Stadium Games!

Since the timeline was tight, the goal was to scale each stadium’s identity

Within the same season, three more stadium partners joined the pilot, including Petco Park and the San Diego Padres.

To move fast, we pulled accent colors from each venue’s logo and populated them across standard UI areas — splash screens, highlights, and key touchpoints — allowing the experience to feel branded without rebuilding the product each time.

Handling Dynamic Stadium Menus

With multiple games in a day, menus often change depending on the event, vendor availability, or venue configuration.

To support this, the system was designed to manage multiple menu configurations under the same venue.

Operators could switch menus quickly for each game without disrupting the checkout flow. This allowed concessions to adapt to different events while keeping the POS experience consistent.

Flow for Creating Menu

Flow for Creating Menu

Design Implementation

1

Determining the right card layouts across cashier, menu, quick menu, and customer screens required multiple rounds of research and testing to keep interactions clear and fast in a high-pressure stadium environment.

Determining the right card layouts across cashier, menu, quick menu, and customer screens required multiple rounds of research and testing to keep interactions clear and fast in a high-pressure stadium environment.

Determining the right card layouts across cashier, menu, quick menu, and customer screens required multiple rounds of research and testing to keep interactions clear and fast in a high-pressure stadium environment.

2

Even with clear swipe cues, some customers tapped or banged their cards on the device, revealing a gap between interface guidance and real-world behavior.

Even with clear swipe cues, some customers tapped or banged their cards on the device, revealing a gap between interface guidance and real-world behavior.

Even with clear swipe cues, some customers tapped or banged their cards on the device, revealing a gap between interface guidance and real-world behavior.

3

Despite a default 15% tip being preselected, many customers switched to custom tip and set it to zero, reducing potential earnings for cashiers.

Despite a default 15% tip being preselected, many customers switched to custom tip and set it to zero, reducing potential earnings for cashiers.

Despite a default 15% tip being preselected, many customers switched to custom tip and set it to zero, reducing potential earnings for cashiers.

Create a free website with Framer, the website builder loved by startups, designers and agencies.