Coles

Retail shrinkage — encompassing theft, self-checkout scan errors, and administrative losses — is a persistent margin drain in large-format supermarket operations. For a network the size of Coles, managing loss prevention manually across hundreds of locations creates both inconsistency and high labor overhead. Traditional approaches rely on staff vigilance and periodic audits, neither of which scales reliably across a geographically dispersed store footprint. At self-checkout terminals specifically, 'skip scanning' — where items pass through without being registered — is a difficult-to-detect revenue leak that compounds across thousands of daily transactions. Without automated, real-time detection at network scale, shrinkage rates remain elevated and uneven, eroding margin at every site.

Coles implemented a layered computer vision system combining two purpose-built AI tools. Skip Scan technology was deployed at self-checkout points, using camera feeds analyzed by AI models trained to detect in real time when items pass through without being scanned. Smart Gates were installed at store exits, applying computer vision to flag potential theft by analyzing customer behavior and basket contents as shoppers leave the premises. Both systems operate on an intelligent edge backbone, enabling on-device inference rather than centralized cloud processing — essential for consistent, low-latency performance across a dispersed store network. Coles cited a partnership with a leading global AI provider as part of this broader operational improvement initiative, positioning the rollout within a wider program of AI-driven efficiency investments disclosed in its FY2024 annual reporting.

The deployment reached significant coverage across Coles' supermarket network within the reporting period:

• Skip Scan active in 546 stores, covering the bulk of the national footprint
• Smart Gates installed in 326 stores, prioritized at higher-risk exit points

Both systems operate continuously, replacing periodic manual audits with real-time automated detection. The scale of rollout — over 500 locations for one system — confirms the technology successfully transitioned from pilot to full network deployment. The shift to automated surveillance also reduces dependence on individual staff vigilance, freeing team members for customer-facing tasks while AI handles continuous loss monitoring.

• Edge AI infrastructure is a prerequisite for large-scale retail computer vision — centralized cloud processing introduces latency and connectivity risk across hundreds of distributed sites.
• Deploying multiple CV layers (exit gates plus checkout scan detection) addresses distinct shrinkage vectors simultaneously, creating more comprehensive coverage than any single solution.
• Differentiated rollout scale (546 stores vs. 326) suggests phased deployment based on store risk profile or infrastructure readiness — a practical model for networks where not all locations present equal risk.
• Partnering with an established global AI provider enables faster enterprise-scale deployment than in-house development, particularly when consistent performance across diverse store formats is required.