Restaurants and cafés: measuring waiting times and table occupancy with cameras

What your takings don’t show

At the end of a service, a restaurant owner knows the takings, the number of receipts and often the number of covers. What they don’t know is how long guests waited at the counter, how many tables sat empty during the rush, or how many people walked away when they saw the queue.

Yet this information matters for everyday decisions: adding staff on certain shifts, rearranging the dining room, opening an extra till, rethinking the terrace layout or the pick-up counter. Video analytics can measure part of this operational reality, without replacing till data.

Dining room and terrace occupancy

The principle is to define zones in the image, per table, per group of tables or per area. Those zones must reflect how the room is actually used. Tables pushed together for a group, a terrace that is reconfigured in summer or a high table shared by strangers all affect the reading, so the zone layout should be agreed with the floor manager and revisited whenever the layout changes.

The system then measures whether each zone is occupied, and for how long. This yields several video-derived indicators:

• dining room and terrace occupancy, hour by hour and day by day;
• average table occupancy time, useful for understanding turnover at different times of day;
• how long a table stays free but uncleared, where the set-up makes that distinction possible;
• comparisons between areas: a packed terrace and a half-empty dining room call for a different decision than a full house everywhere.

Counter, pick-up and drive-thru queues

For quick-service restaurants, coffee shops and takeaway outlets, the queue is often the most telling indicator. A zone in front of the ordering counter estimates how many people are waiting and the average wait. A zone at the pick-up point measures how long people wait for their order.

For a drive-thru, counting vehicles and timing them from lane entry to exit gives an estimate of total visit time. These measurements are about flows, not identified individuals: the point is to know when the queue builds up, not who is in it.

Alerts can be set to trigger when a threshold is crossed, for example more than six people at the counter for several minutes, so the floor manager can open another till. The threshold should be fine-tuned during the pilot to avoid a flood of notifications.

Camera placement and privacy masks

Security cameras usually point at the till or the entrance. To measure table occupancy, a high vantage point covering the dining room, without too many obstructions from pillars, plants or partitions, generally works better. Terraces bring their own challenges: direct sunlight, parasols and big swings in light between lunchtime and evening.

Privacy masks matter here. They exclude the kitchen and workstations, the street visible from the terrace and anything unrelated to the measurement. On-site processing (edge AI) keeps footage from travelling: only counts and durations reach the dashboard.

The audit checks whether existing cameras offer a usable angle, whether reframing is enough or a dedicated camera is needed, as well as stream access and image quality in the evening.

Linking with covers and receipts

Video indicators become more useful when set against till data. Comparing dining room occupancy with the number of covers, or queue length with the number of receipts in the same time slot, helps you see whether a dip in sales comes from low footfall or from service capacity running at its limit.

This requires an integration with the point-of-sale (POS) system, which depends on its export or API options and calls for a dedicated study. A few caveats apply: an occupied table does not always mean one receipt, a group may pay separately, and timestamps must be aligned. On the dashboard, each indicator should show its source: video for occupancy and waiting, the till for sales and covers.

Starting with a single-venue pilot

Before equipping several venues, a pilot on one representative site shows what the cameras can actually measure. It should run long enough to cover lunch and dinner services, weekdays and weekends, and the changing light on the terrace. A typical approach:

• pick two or three precise questions, for example: at which times does the counter wait exceed an acceptable level, and is the terrace full while the dining room is not?
• audit the cameras, adjust framing and set up privacy masks;
• check the reliability of occupancy and queue figures against manual spot counts;
• fine-tune alert thresholds based on the notifications the floor manager genuinely finds useful;
• decide, based on observed results, whether to extend, adjust or stop.

Informing guests and staff

The restaurant operator is the data controller under Law 09-08. It is their responsibility to inform guests and employees with visible signage, complete any applicable formalities with the CNDP and set limited retention periods. When an analytics purpose is added to existing security cameras, staff should be informed again.

For teams, the message needs to be clear: the system measures occupancy and waiting times to organise service better; it is not there to monitor or individually assess waiters or kitchen staff. Analyses that focus on flows are easier to justify and more readily accepted than those targeting people. No facial recognition or identification is needed for any of these measurements.

Want to measure waiting times and occupancy in your venue? Take a look at our restaurants and cafés page, then request an audit of your cameras to prepare a single-site pilot.