Computer Vision Based
Self Checkout
TL;DR
Designed a computer vision-powered self-checkout experience that improved checkout efficiency and scaled to 2,000+ stores.
ROLE
Lead UX Designer
Realized
May 2024
PLATFORM
Tablet
Background
Shaping the future of convenience
7-Eleven is a global convenience store chain known for its wide range of products and innovative shopping experiences. With a presence in over 17 countries, it aims to provide quick and easy solutions for customers' daily needs.
QCO
A quick self-checkout system to reduce wait times and streamline the checkout process.
Impact of Design
2,000
Stores with Self-Checkout
$47 million
Sales Generated
$48 million
Saved in Annual Labor Costs
7 Million
Transactions Processed
My Role
Visual Execution
Created high-fidelity mockups and final visual designs.
Prototypes
Built interactive prototypes for testing and validation.
User Testing
Conducted sessions and iterated based on feedback.
Why do customers use checkout?
Insights
54% of users found the self-checkout interface confusing, leading to errors and delays
85% had issues with item scanning, like items not scanning correctly or needing multiple attempts.
67% faced problems with the payment process.
User Journey
Challenges
Accuracy and Reliability
Ensuring the computer vision system accurately identifies and classifies items is crucial. To mitigate recognition errors, we need to provide users with clear guidance on using the system correctly, like how to position items.
Effective Item Searches
When users manually input items, they often face challenges with search results that do not match their terms accurately. Common issues include misspellings, similar item names, and insufficient suggestions. We need to enhance search functionality to improve accuracy and relevance of search results.
Problem Reporting
Since the algorithm isn't fully accurate, issues like unrecognized or similar items may arise. We aim to provide simple, user-friendly solutions to help users quickly resolve these problems.
Ideation
In our ideation process, we began with group brainstorming sessions to collaboratively generate and conceptualize ideas. These sessions fostered creativity and allowed us to explore a wide range of possibilities. After narrowing down the best concepts, we created low-fidelity prototypes to test and validate these ideas. These prototypes were essential in identifying potential issues and gathering initial feedback, which informed subsequent iterations and refinements.
Happy Path
"What's this?"
In addition to the main user flow, a key feature is the "What's This?" option.
If the algorithm fails to recognize similar items or items within a container, users are prompted with "What's This?" to allow them to manually identify the items.
We've developed dedicated flows to address these edge cases.
A/B Test
Methods
Moderated usability tests – 10 self-checkout customers
Data collected: January 24-26th
Goals
Understand delights, pain points and opportunities
Understand the usability of QCO unrecognized item flow
A/B Test Summary
01
CTA hierarchy is important.
If Finalize & Pay is primary action, then that CTA should be prioritized or highlighted.
02
Customers prefer clear instructional prompts. A full-screen experience for adding items is favored as it reduces distractions and has clear guidance.
03
Users mentioned that the graphics displayed provided instructional information and they were pleased with the look & feel.
Final Design
Usability Study
Observe store customers as they use the QCO and note any technical or usability issues they encounter, using a pilot QCO at the Sound 7-Eleven Store in Coppel, TX.
60+
Store customers
6 hours
Conducted during 6 hours
Summary
Almost everyone was willing to try the QCO
Associates were able to spend more time cleaning, stocking shelves, and preparing fresh food when customers were willing to use self-checkout without assistance.
However, a few customers stated an overall preference to bechecked out by an Associate.
“
"That's as easy as it gets"
"Wow that's phenomenal"
"Amazing"
"How awesome"
"Pretty easy"
47 store customers who used the QCO rated its ease of use
“
Key Metrics
Following the usability study, we established specific metrics to measure the success and impact of the self-checkout system.
Transaction Goal:
15%
eligible transactions to be processed through the self-checkout system.
Transaction Fit:
<40%
of transactions that are suitable for self-checkout are diverted to traditional registers when the self-checkout system could have met their needs.
Customer Retention:
50%
of customers who use the self-checkout system for the first time return to use it again.
Search Efficiency:
80%
success rate for customers using the search function for items
By setting these metrics, we aimed to ensure that the self-checkout system not only met usability standards but also achieved business goals related to customer satisfaction. Monitoring these KPIs allowed us to make data-driven decisions and further refine the system based on real-world usage and feedback.
