Skip to content
UA Discuss your project
  1. Home
  2. Resources
  3. Checkout queues: how to count and shorten them with cameras

Checkout queues: how to count and shorten them with cameras

5 min read · Updated 2026-10-28

Author: DuoVision team

For: managers of shops and chains, operations directors in retail and petrol stations, heads of administrative service centres (CNAP) and reception offices

Summary

  • A camera above the checkout area counts how many people stand in a defined queue zone; the system does not identify who they are and does not recognise faces.
  • Waiting time is estimated indirectly: the average queue length is divided by the service rate taken from POS receipts.
  • When the queue stays above the threshold for longer than a set time, the shift supervisor receives an alert to open another till, and a report by hour and day of week shows when tills are short.
  • A dense queue, overlapping people and a low camera angle reduce accuracy, so we set the thresholds for each shop and first run the system in shadow mode.

For: managers of shops and chains, operations directors in retail and petrol stations, heads of administrative service centres (CNAP) and reception offices

What exactly the camera counts

In the frame of a camera that looks at the checkout area, we define one or more queue zones. Usually this is the lane in front of each till, or a shared waiting zone if the customer chooses the till. Several times a minute the system counts how many people are in each zone. The cashier behind the counter and the person already paying can be excluded with a separate zone so that they are not added to the queue.

The result is a simple series of numbers: the number of people in the queue for each till at each moment. Both alerts and reports are built from it. The system does not follow individual people between cameras and does not recognise faces: to count, it does not need to know who is in the queue.

Waiting time as an estimate

The most accurate waiting time would come from tracking each person from joining the queue to the till. In practice this is hard and unnecessary: in a dense queue people overlap, move to another till and come back for a forgotten item. So we estimate waiting time indirectly.

We take the average number of people in the queue over a few minutes and divide it by the service rate, that is, the number of people the till serves per minute. The rate is best taken from the receipts of the till system. For example, if on average six people stand in the queue and the till rings up two receipts a minute, the last person in the queue waits about three minutes. For the decision "open a till or not" this estimate is enough, and it does not require following a specific person.

The "open another till" alert

The alert fires when the queue stays above a threshold for longer than a set time. You set the threshold and the duration yourself, for example "more than four people per open till for three minutes". Short spikes, when a group of people arrives at once, do not trigger the alert.

The alert goes to whoever can act: the shift supervisor, the floor administrator, the duty manager. It arrives as a card with a frame of the queue, so the person immediately sees whether the queue is real or whether staff with a trolley are simply standing by the till. When the queue drops below a second, lower threshold, the system records that the situation is closed, and the time between the alert and normalisation goes into the report.

The alert delivery channel, for example a messenger or a panel on the administrator's monitor, is set up for your site.

An hourly report for the shift schedule

The alert is for reacting now, and the report helps plan the next week. After a few weeks of work a picture builds up by hour and day of week. It shows when the queue grows regularly and when open tills stand idle.

Report metricWhere it comes fromWhat it is for
Average and maximum queue length per hourCameraShows the hours when tills are short
Number of "open a till" alerts and time to normalisationCamera and card logShows how quickly the shift reacts
Number of open tillsTill system or a camera above the tillsCompares load with open tills
Service ratePOS receiptsEstimate of waiting time
Estimated waiting timeCamera and POS receiptsComparing days and shops with each other

This report can be put next to the shift schedule to see where people are placed at the tills out of habit and where by load. If you also need to see how many staff are actually in each area during a shift, that is a separate task, described on the page about staff counting.

Cross-checking with POS receipts

The camera gives the number of people, the till system gives the number of receipts. Together they answer questions that cannot be answered separately. If a counter camera stands at the entrance, the number of receipts divided by the number of people who came in gives the conversion of visitors into buyers by hour. If the queue is long and receipts are few, the problem is service speed: a slow terminal, age checks, returns.

We build the integration with the till system for the task: at the survey we find out what data it provides and with what delay. More scenarios for shops are collected on the page about retail, and for petrol stations, where queues form both at the pumps and at the till, on the page about petrol stations.

Where the camera sees poorly

  • When people stand very close together, a side-mounted camera sees two people as one silhouette, and the count comes out low.
  • Racks with small goods near the tills, advertising stands, trolleys and columns hide part of the queue.
  • A camera that looks from above or at a steep angle sees heads separately and counts most accurately. A camera at eye level pointed into the depth of the hall suits less well.
  • The far end of a long queue on a wide shot takes up a few dozen pixels in the frame and may drop out of the count.
  • A cashier, a guard or a shelf stacker in the queue zone is added to the count unless excluded with a separate zone.

So we first run the system in shadow mode: for a few weeks it only counts, and we compare its count with what a person sees on the same frames. We switch on alerts to staff after this check.

Customer privacy

For queues, anonymous counting is enough. The system stores numbers: how many people, in which zone, at what time. Frames are needed only for alert cards and are kept for as long as you set in your video surveillance policy.

The video recording of customers itself remains processing of personal data if a person can be recognised in it. Under Article 12 of the Law "On Personal Data Protection", people must be informed about processing when data is collected, so a video surveillance sign at the entrance is mandatory. Under Article 6, data in a form that allows a person to be identified is kept no longer than needed for the purpose. For more, see our guide on video analytics and personal data.

The same for CNAP and reception offices

In administrative service centres (CNAP), reception offices and branches with an electronic queue, the waiting hall works by the same logic. The camera counts people in the hall and at the service windows, the system sends the administrator an alert when there are too many people, and the hourly report shows on which days it is worth opening extra windows. If the electronic queue issues tickets, the number of tickets plays the role of POS receipts. Scenarios for state and municipal bodies are described on the page about the public sector.

Where to start

  1. Take POS receipts for a few weeks and see at which hours and days of the week there are the most receipts per open till.
  2. Check whether your existing cameras see the whole checkout area from above or at a steep angle, and whether racks and stands hide the queue.
  3. Define the threshold: how many people per till and how many minutes are already too many for your shop, and who on the shift receives the alert.
  4. Ask the supplier whether you can start with shadow mode on your cameras and how the system excludes cashiers and other staff from the count.

We connect to the cameras that are already installed and start with shadow mode; you can write to us through the contacts page.

Where to apply this

Sources

  1. Law of Ukraine «Про захист персональних даних» (“On Personal Data Protection”) of 01.06.2010 No. 2297-VI, Articles 6 and 12

Let’s discuss your site

Request a consultation or a project estimate. We will tell you what your cameras can monitor, where it is best to process the video and how much it will cost.