A video analytics system can send an alert within seconds. But an alert on its own does nothing to reduce shrinkage. What makes the difference is what the team does in the minute that follows: who receives it, who reviews it, who goes to the spot, and how. Without clear rules, the same problems keep coming back. Notifications ignored because "someone else is dealing with it." Clumsy interventions that put an employee in a difficult position. Or, conversely, disproportionate reactions to a gesture that was perfectly normal.

This guide is intended for store managers, security managers and integrators supporting a deployment. It offers a simple method for building an alert response protocol, adapted to the size of the store and its staffing.

Why an alert is only as valuable as the response that follows

Detection tools are advancing quickly, organizations less so. Yet technology only covers part of the journey: it flags an at-risk gesture (a concealed item, a product slipped into a personal bag, a package opened in the aisle, consumption on the spot) and sends a video clip. What happens next is entirely down to people. This is the principle behind Oxania: technology detects, the team decides what happens next.

The Loss Prevention Research Council, which evaluates prevention solutions through controlled trials and interviews with offenders, structures its research in three stages: influencing choices upstream (signage, lighting, cameras), detecting, then sharing the facts and information gathered with law enforcement, local stakeholders and other retailers (see the LPRC research method). Detection is only one link. The response protocol is precisely what connects it to the rest of the chain.

The financial stakes are real. According to the latest total published by the National Retail Federation, shrinkage accounted for $112.1 billion in losses for US retail in 2022, up from $93.9 billion the previous year (NRF press release on 2022 shrinkage). An alert that goes unhandled is a missed opportunity to limit a loss.

Defining roles: who receives, who verifies, who acts

The first question is not technical, it is organizational. When everyone receives everything, no one feels responsible. It is better to distribute three roles, even if one person holds several of them in a small store:

  • The recipient receives the notification. Depending on the time slot, this is an aisle manager, a security officer or the store manager. The device matters little (mobile, tablet, computer), as long as it is in the hands of someone available and not forgotten in an office.
  • The verifier watches the clip and decides whether the situation warrants a presence in the aisle. This is often the recipient themselves, but the role must be named.
  • The responder goes to the zone concerned, with a clear instruction and limits known to everyone.

The recipients' schedule must follow that of the teams. An 8 a.m. opening with two people is not handled like a Saturday afternoon with one agent on duty. Statistics by zone and by time slot help align this schedule with the genuinely exposed moments rather than with impressions. When the store has private security, the allocation should be planned with them: our article on how AI and security agents complement each other details this division of tasks.

Three levels of response, from the most low-key to the most formal

Not every alert calls for the same reaction. An effective protocol grades the responses to avoid two opposite pitfalls: indifference and escalation.

  1. Level 1, presence. A team member goes to the aisle, tidies a shelf, offers help. This is the default response: the person who was hesitating realizes the aisle is staffed, and a genuine customer is simply welcomed.
  2. Level 2, the courteous exchange. If the clip clearly shows an item placed in a bag or under clothing, a trained team member can, at the checkout or before the exit, politely remind the person that an item remains to be paid for. Never an accusation, never physical contact, and always within the framework of the retailer's policy and local law.
  3. Level 3, preserving the evidence. When the situation is confirmed and the loss is real, the clip is retained and can be used for reporting to the authorities, if management so decides. This step falls to the manager, not the responder.

Verification speed determines this choice. With Oxania, the video alert arrives in under 10 seconds and an automatic progressive zoom is applied to the point of interest: the verifier assesses the gesture in moments, without going through minutes of footage.

Team safety comes before merchandise

This is the rule that should appear at the top of any protocol. In its 2026 study on theft and violence in US retail, the NRF reports that 40% of the retailers surveyed see more aggression and violence during shoplifting incidents than a year earlier (it was 73% in the 2024 edition), and that 81% say thefts involving violence have led them to strengthen their employees' training on workplace violence (NRF 2026 study). For a store manager, this translates into a few simple guidelines, to be posted in the stockroom:

  • never chase a person outside the store;
  • never intervene alone if the situation becomes tense, and immediately alert the manager on duty;
  • step back as soon as an exchange turns hostile, leaving it to the video to preserve the facts.

No item is worth an injured team member. A protocol that states this in black and white also reassures teams, who are more willing to adopt a tool whose consequences they do not fear.

Qualify, measure, adjust

Each handled alert should receive a simple status: confirmed, no action taken or not relevant. These annotations feed a monthly review built around a few indicators:

  • the average time between the alert and its review;
  • the share of alerts actually handled;
  • the breakdown of confirmed alerts by zone and by time slot;
  • the aisles where a level 1 presence is enough, and those where it is not.

These figures build on the standard shrinkage indicators presented in our article on the essential KPIs for anticipating losses. They are also useful to the systems integrator: a high rate of irrelevant alerts in a zone can indicate a camera angle that needs to be reviewed, a topic covered in the role of integrators in detection effectiveness.

Training and keeping the protocol alive

A protocol filed away in a binder protects no one. Team engagement goes hand in hand with lower losses: an ECR Retail Loss study, based on questionnaires from more than 200,000 employees of three large European retailers (2012 data), estimates that if the quarter of stores with the lowest engagement scores reached the average level of the other stores, shrinkage in those stores would fall by 12.5% (ECR report on employee engagement and shrinkage).

In practice, three habits are enough: a one-page quick-reference sheet given to every new arrival, seasonal staff included; a short role-play based on real clips during team meetings; and a monthly review of alerts that highlights good reactions rather than pointing out oversights. It is through this regularity that the tool becomes a collective reflex, and that the reduction of shrinkage takes hold over the long term.