Sam's Club members scan items on their phone and walk out without ever touching a register. I was the solo UX designer on the features that handled real complexity at scale — multi-transaction support and shrink reduction — contributing to a product that now drives 1M+ monthly transactions and an NPS of 85.
Designs shown reflect the version I shipped in 2022–2023. The product has since been updated with a visual rebrand; core interactions remain the same.
Members scan items as they shop, pay in the app, and walk out. No cashier, no kiosk, no line. The core experience was already live when I joined — my work focused on the features that handled real complexity as adoption scaled.
I was the sole designer on this feature, working directly with a dedicated research team and engineering. The happy path direction was validated quickly — the real work was designing for the edge cases that stakeholders initially deprioritized. I pushed for full edge case coverage before launch, which required close collaboration with engineering to resolve technically complex scenarios.
10 of 12 participants noticed the "Purchase more items" button without prompting. The entry point wasn't the problem — the anxiety was about what came next. Would they have to rescan everything? What are the limits? Would it affect their first purchase?
A "Before you purchase more items" modal sets expectations upfront — spend limit, separate receipt — so members can make an informed choice before committing to a second scan.
Most participants found the persistent banner — "You already paid for 5 items on a separate receipt" — genuinely reassuring. It confirmed their first cart was safe and made starting the second feel intentional, not risky.
Every participant correctly identified that the single QR code covered both purchases and knew to show it once at exit. However, they relied on the item count as their primary signal — which raised a concern: with a large cart, it would be hard to mentally verify the total across two separate transactions. The receipt design addressed this by showing the most recent purchase at the top, which aligned with what members naturally expected.
Stakeholders initially wanted to ship the happy path and address edge cases post-launch. I pushed for full coverage before release — a broken edge case at the exit arch in a live club is a far worse member experience than a delayed launch. Two cases in particular required deep collaboration with engineering to resolve.
Age verification is tied to transaction state — if the first cart had no alcohol, the age gate wasn't triggered. When a member adds wine in the second scan, the system has no record of ID verification. Engineering and legal both flagged this as a blocker. The flow required a mid-session age verification step that didn't interrupt or confuse the member about which transaction they were in.
If the second transaction triggers a fraud flag, the member has already exited Scan & Go once. The system can't simply fail — the member is mid-club, the first payment is complete, and the items from the second scan are in their cart. The flow had to handle transfer to POS, preserve the first receipt, and give the member a clear path forward without losing their cart or causing double-payment anxiety.
I led the UX strategy for enabling up to two Scan & Go transactions in a single club visit — solving a major pain point for the 10% of members who regularly make multiple purchases. Through iterative research, usability testing, and deep collaboration with engineering to resolve complex edge cases, the shipped experience delivers a seamless second-transaction flow at scale.
Reduce Item Fail Rate by 20% on Scan & Go. (Item Fail Rate = of all items scanned at exit, how many were flagged as not paid.)
Once members tap checkout, they're mentally done scanning. Concepts that pulled them backward — confirming a count (V2) or manually entering item numbers (V3) — caused real frustration. V3 prompted strong reactions: "This would make me stop using Scan and Go." V1 (passive nudge) rated ~8.5/10. V3 rated 1.2/10. Research validated the low-friction direction — but member preference alone couldn't determine how much friction was enough.
The winning variant surfaces an item count modal when the member initiates checkout — a quick confirmation of how many items were scanned before they proceed. No extra action required for honest members. Low enough friction that it barely registered, effective enough to move the metric.
−16.4% Item Fail Rate · NPS held at 85
Final shipped version exceeded the 20% target.
