AI Insights & Surveillance Platform

A 0-->1, real-time theft tracking platform to help major retailers reduce theft in stores. AI camera systems were watching self-checkout lanes across major retailers in real time, but clients had no way to determine if the cameras were actually saving money. I led design and research end-to-end that turned that raw AI detection into a self-serve dashboard retailers could actually act on.

$9.5M

recovered / client / yr

97.5%

increase in data accessibility

43%

productivity increase / week

$730K

SaaS revenue / customer / yr

$3.2M

revenue, yr 1

40K

lanes implemented

My Role

Design & Research Lead · 0→1

Team + Stakeholders

PM · PO · 3 Designers · 5+ Customer Success Managers

Problem

Retailers couldn’t tell if Toshiba’s AI-powered camera system was paying off.

Toshiba's AI camera systems were cameras watching self-checkout lanes across major retailers in real time, flagging theft and shrink as it happened. Information about the system was only accessible via a Toshiba employee who manually translated reports for each client, meaning retailers could only see or understand their data every few weeks. There was no portal or self-serve access, but even if there was, users would not be able to interpret the charts. The charts themselves were poorly displayed and difficult to read, and were also intended for internal debugging of the system, not to display ROI to customers.

The metrics needed to speak to the people running stores day to day, not the developers who built the underlying system. I needed to prove the system saved money and reduced theft.

Solution

New data visualizations, visibility intro trends, direct access to actionable data, and issues surfaced quickly.

01

AI recognizes and categorizes theft events into trends, using video as supporting evidence.

The initial product vision was a tool for reviewing footage clip by clip, but that wasn't what users were asking for. They wanted to see broad trends and then be able to drill into the data as needed.

Video became supporting evidence attached to flagged events, not the centerpiece. AI pre-sorts what matters so managers aren't watching hours of footage.

02

ROI is surfaced immediately across four savings categories, prioritizing what’s most important

I analyzed available model data to translate 53 charts into system performance, time savings, sales/overall savings, and loss events. Every number is backed by data the user can click into.

Clients can now easily interpret the amount the system has saved them.

03

Users can focus on decisions, not noise.

AI sorts through massive amounts of data to surface what users need most. Every layer is now accessible to users, retained, and surfaced in real time.

HQ can now see and understand every store's performance and the product’s savings impact easily.

Impact

For the first time, users could now access and understand real-time data

Implemented in over 40,000 retail lanes, there was $9.5M theft recovery per client per year, 97.5% increase in data accessibility, and 43% higher internal productivity for Customer Success Managers — turning a manual, weeks-long process into real-time, self-serve ROI.

$9.5M

recovered / client / yr

97.5%

increase in data accessibility

43%

productivity increase / week

$730K

SaaS revenue / customer / yr

$3.2M

revenue, yr 1

40K

lanes implemented