Measuring wait times at airport security checkpoints involves continuously estimating the time between when a passenger enters the security checkpoint line and when they actually pass through. Three types of methods are used: manual recording by staff, dedicated sensors installed on-site, and video analysis using existing cameras. The latter is gradually becoming the standard at major hubs because it requires no new equipment and covers the entire waiting area, rather than a single checkpoint.
Reducing wait times, however, does not depend on the measurement itself but on what is done with it: opening an additional lane before the line becomes saturated, redeploying agents from one PIF to another, and informing passengers in advance. The purpose of the measure is to provide a sufficiently early and reliable signal so that these decisions can be made before the peak, rather than observed after the fact.
Why is wait time at border checkpoints so difficult to measure?
A border inspection point (BIP) line is almost never a straight line. It winds, splits into several lines, absorbs priority lines, and is constantly reorganizing. A counter placed at the entrance therefore tells us nothing about the actual time spent inside.
Added to this is volatility: the number of people at a border inspection point can double in about ten minutes as waves of travelers depart, then drop just as quickly. A count taken every hour by an officer provides a snapshot at a specific moment, never the trend over time.
Finally, metrics reported retrospectively—in a weekly or monthly report—arrive too late to be actionable. They’re used to determine staffing levels for the following quarter, not to prevent congestion on Tuesday mornings.
What methods are available for measuring wait times at PIFs?
Manual Counting
An agent times the journey of a test passenger or counts the number of people present at regular intervals. This is simple to implement and useful as a baseline for monitoring, but the sample size remains small, and the method ties up agent time that is needed elsewhere.
Dedicated sensors
Traffic sensors, connected walkways, satisfaction kiosks, and detection of Wi-Fi or Bluetooth signals from passengers’ devices: these devices provide continuous measurement but require physical installation, dedicated maintenance, and coverage limited to equipped locations. Furthermore, radio signal detection depends on what the passenger has enabled on their phone, which introduces a variable bias.
Video analysis using existing cameras
In virtually all airports, the waiting area at a border control checkpoint is already covered by video surveillance cameras. Video analysis using artificial intelligence leverages these feeds to generate occupancy and traffic flow metrics without adding any equipment to the area.
How does video analysis calculate wait times?
The principle is based on two complementary metrics, continuously generated for defined areas of interest within the image:
- density: the number of people in the line at a given moment;
- throughput: the number of people passing through the exit checkpoint per minute.
The estimated wait time is derived from these two values: a line of 120 people in front of a checkpoint that processes 12 passengers per minute corresponds to a wait time of approximately ten minutes. This is a direct application of Little’s law, which relates the average number of elements in a system to their arrival rate and their residence time.
Anonymous tracking of movement patterns then refines the estimate by distinguishing between lines that are moving forward and those that are stagnating, and by identifying backflows toward upstream areas. No personal identification is involved in this calculation: these are silhouettes counted and tracked over a few meters, not recognized individuals.
What conditions are necessary for the measurement to be usable?
Three factors determine the reliability of the result far more than the raw performance of the models.
The camera angle. A camera positioned too low causes silhouettes to overlap and undercounts dense lines. Bird’s-eye views, even partial ones, yield significantly more stable measurements. It is rarely necessary to reposition the entire array—often one or two cameras are sufficient to correct an area.
Zone delineation. A poorly defined queue mixes waiting passengers with those moving through the concourse. Thezone delineation must align with the actual layout of the passenger check-in area and be revised when that layout changes—which happens during peak periods.
Operational integration. A metric that sits in a dashboard that no one opens does nothing to reduce wait times. Alert thresholds must be escalated to where decisions are made: the supervision station, the zone manager’s screen, or a mobile notification.
How do we move from measurement to reducing wait times?
Measurement becomes useful when it triggers an action at a specific moment. Three approaches underpin most deployments.
Early warning. A density threshold, set below the saturation point, allows time to open an additional lane or call for reinforcements. It is the timeliness of the signal that creates value, not its precision down to the individual passenger.
Redistribution between zones. Comparing the workload of multiple border checkpoints in real time allows agents to be shifted to the one experiencing an increase, rather than sizing each station for its own peak.
Retrospective planning. Cumulative curves by day and by hour provide a factual basis for scheduling shifts, making decisions on facility improvements, or objectively discussing departure slots with airlines.
Is this measure compliant with the GDPR?
It can be, provided it remains within a specific framework. The CNIL has published guidelines on enhanced cameras in public spaces, which clearly distinguish between image analysis for statistical purposes and the identification of individuals.
A compliant approach relies on anonymous processing: no facial recognition, no biometric templates, and no re-identification of a passenger across cameras. The images are analyzed in real time, and only aggregated data—counts, density, estimated time—are retained.
The standard requirements remain: informing the public about the affected area, registration in the data processing registry, and an impact assessment when the context requires it. These elements are evaluated on a case-by-case basis: the operator’s DPO remains the point of contact who makes the final decision.
Where to start?
A deployment is rarely tested on the scale of an entire terminal. The most common approach is to select a pilot PIF covered by two to four existing cameras, define a single operational threshold with the operations teams, and then compare the generated values with the usual manual readings over the course of a few days. This calibration phase is more valuable than any demonstration: it builds trust in the data, without which no alert will be acted upon.
XXII is developing CORE, a French video analytics platform that leverages existing cameras to generate these types of metrics (density, flow, estimated wait times, threshold alerts) without facial recognition and in compliance with the GDPR. This approach applies to airport security checks as well as other checkpoints within a transportation hub.