SAI Groups

Aisle Monitoring

Detailed view

Retail shrinkage increasingly occurs inside the store, long before a customer reaches the checkout. Traditional surveillance systems rely heavily on manual monitoring and post‑event investigation, making them reactive and inefficient. SAI Group’s Visual AI Platform addresses this challenge through its Aisle Monitoring feature, which uses behaviour analysis and product‑counting intelligence to detect concealment, walkouts, counterflow exits, and partial payments in real time.

The platform continuously monitors customer activity across aisles and correlates it with checkout and exit behaviour when those cameras are available. It detects suspicious actions such as concealment based on behaviour, picking items without presenting them at checkout, walking out the way the customer entered (counterflow), and paying for only a subset of picked items. Alerts are generated in under 6-8 seconds, complete with images, video snippets, and the aisle name where the alert was triggered—enabling store staff to respond quickly and effectively.

Why the Aisle Monitoring Feature Is Important for Retail Stores

Shrinkage is no longer limited to overt theft. Modern retail loss often involves subtle behaviours such as item concealment, delayed concealment across different aisles, walkouts without checkout interaction, or partial payment at self‑checkout counters. These behaviours are difficult to detect using rule‑based systems or human observation alone.

The Aisle Monitoring feature is critical because it:

  • Shifts loss prevention from reactive review to real‑time intervention
  • Identifies suspicious behaviour even when theft is distributed across multiple aisles
  • Bridges the visibility gap between product selection and checkout behaviour
  • Reduces dependency on continuous manual monitoring of CCTV feeds
  • Enables consistent detection across stores, formats, and staffing levels

    • By focusing on how customers behave, not just where they are, the feature helps retailers prevent losses before they materialise.

    How the Aisle Monitoring Feature Works

    Behaviour‑Based Concealment Detection (Aisle Cameras Only)

    When cameras are deployed in aisles, the platform analyses customer behaviour to detect concealment events. Concealment is identified based on actions and movements rather than simple object disappearance.

    Key capabilities include:

  • Detecting when a customer conceals an item in the same aisle
  • Detecting when a customer picks items in one aisle and conceals them later in another aisle
  • Tracking customer movement across multiple camera views to maintain behavioural continuity

    • If only aisle cameras are installed, the platform will generate alerts specifically for concealment events.

    Counting‑Based Walkout and Partial Payment Detection

    (Aisle + Checkout / Exit Cameras Required)

    When checkout and exit cameras are added, the platform correlates aisle activity with checkout behaviour using product counting intelligence.

    This enables detection of additional loss scenarios, including:

    • •Walkout: Items are picked up, but the customer exits without going to checkout
    • •Counterflow Walkout: The customer exits the store the same way they entered, bypassing checkout
    • •Partial Payment: The customer takes multiple items but pays only for a subset at checkout
    • •Non‑Presentation: Items are picked but never presented at checkout, even without visible concealment

    Without checkout or exit cameras, these scenarios cannot be confirmed. With full camera coverage, the platform links what was picked to what was paid for or not paid for.

    Real‑Time Alerts and Evidence

    Once suspicious activity is confirmed, the platform sends an alert in under 8 seconds. Each alert includes:

    • • A clear image of the individual involved
    • • A short video clip capturing the relevant behaviour
    • • The aisle name where the person is located at the time the alert is generated

    This allows store teams to respond while the customer is still on the premises, rather than reviewing footage after the loss has occurred.

    Benefits of Using the Aisle Monitoring Feature

    Proactive Loss Prevention

    Real‑time alerts enable staff to intervene before a customer exits the store, reducing shrinkage rather than simply documenting it.

    Coverage Across the Entire Shopping Journey

    By linking aisle activity with checkout and exit behaviour, the platform provides end‑to‑end visibility—from product selection to store exit.

    Detection Beyond Simple Concealment

    Even when customers do not visibly conceal items, the system can detect losses caused by walkouts, counterflow exits, or partial payments.

    Reduced Operational Burden

    Automated detection reduces reliance on manual CCTV monitoring and subjective judgement by store personnel.

    Actionable Alerts

    Alerts include visual evidence and precise location context, allowing staff to act confidently and appropriately.

    Scalable and Store‑Ready

    The feature adapts to different store layouts and camera deployments, delivering value even with aisle‑only installations and enhanced capability with full coverage.

    FAQ

    What can the platform detect with only aisle cameras installed?

    With aisle cameras alone, the platform can detect behaviour‑based concealment, including concealment occurring in a different aisle from where items were picked.

    What additional detections require checkout or exit cameras?

    Checkout and exit cameras are required to detect walkouts, counterflow walkouts, partial payments, and non‑presentation of items at checkout.

    Can the system track a customer across multiple cameras?

    Yes. The platform can detect when a customer picks items in one camera view and conceals them in another.

    How quickly are alerts generated?

    Alerts are generated in under 8 seconds, enabling near real‑time response.

    What information is included in an alert?

    Each alert includes a picture, a video clip, and the aisle name where the person is located at the time of alert generation.

    Does the platform rely on fixed rules?

    No. Detection is driven by behavioural analysis and counting logic, making it effective even when theft patterns vary.