GSX 2026: Video Analytics Questions for Security Managers

GSX 2026 runs September 14–16, 2026 at the Georgia World Congress Center in Atlanta. Hosted by ASIS International, it draws security management professionals from more than 80 countries. Confirm dates and registration with ASIS directly.

GSX has a different centre of gravity from the integrator shows. The people on the floor own programmes, budgets and risk registers rather than installs — which changes what is worth asking about analytics, because the hard problems at that level are rarely technical.

Ask about the alerting burden, not the detection rate

Detection accuracy is the number every vendor leads with and rarely the constraint that decides whether a deployment succeeds. The constraint is what lands on a person, how often, and whether they can act on it.

Worth asking: what is the expected alert volume per camera per night in the first fortnight, and what does it look like after tuning? A vendor with real deployments knows both numbers and knows the first is uncomfortable. One who quotes a single low figure has either not deployed at scale or is describing a system with the sensitivity turned down, which quietly removes the detections you wanted.

Ask where detection ends and response begins

This is the central confusion in the category and it matters more at programme level than anywhere else, because the two are different purchases with different obligations, different insurance implications, and different failure modes.

• Analytics detect, classify and surface. The action is yours.
• Monitoring puts a person in a centre watching. That is a service, with staffing and continuity obligations behind it.
• Response dispatches or intervenes. Different again, with a different liability profile.

Plenty of vendors sell more than one. Be clear which you are buying and what is contractually promised, because "24/7" appears in marketing for all three and means something different in each. For the record: we are the first of those. VuePoint's platform detects, classifies and hands to a person — the escalation decision stays with the operator who owns the site.

Ask what the evidence record looks like eighteen months later

At management level, the value of a detection is often realised long after the event — in a disputed claim, an investigation, or a regulatory conversation. That makes retention, export and chain-of-custody questions more important than they feel on the floor.

• How long is footage retained by default, and what does changing that cost?
• Can you export the original file, not a re-encoded clip?
• Is there a defensible audit trail of who viewed what, and when?
• If you leave the vendor, what happens to the historical event record?

That last one catches people. The footage is usually exportable; the structured event history — the thing that makes a pattern visible across eighteen months — frequently is not.

Ask about the privacy position, in writing

The same applies to employee monitoring, which carries notification obligations that vary by jurisdiction. Worth your own counsel rather than a vendor's assurance.

The programme question nobody at a stand will raise

Who acts on the alerts, at what hours, and what happens when they are on leave? Detection is rarely the constraint in a mature programme — the response capacity is. Adding a better detector to a queue nobody has time to work changes the vendor on the invoice and nothing else.

It is worth walking GSX with a realistic count of how many events per week your organisation can genuinely action, and treating that as the design constraint. It usually rules out more options than any technical requirement.