Operations
Video analytics for restaurants — six numbers your cameras already know
If your cameras are only used for after-the-fact security, you're paying for half a system. Six restaurant KPIs your existing camera footage already knows.
Most restaurants we walk into already have camera coverage that would shock them if they could see what it knows. Modern IP cameras — even mid-range ones from 2020 — have onboard analytics or cloud-side processing that can answer operational questions traditional metrics can't.
Below are six numbers we routinely pull from existing camera setups. None of them require new hardware in most cases. All of them quietly answer questions that operators have been guessing at.
1. Door entry count by daypart
Most POS reports tell you tickets per daypart. They do not tell you walk-ins per daypart. The gap between the two is your conversion rate — how many people walked in and didn't buy. We've seen restaurants where the lunch conversion was 88% and the dinner conversion was 41% — the same brand, same staff, different decisions in the dining room. You can't fix what you can't see.
2. Average time-in-line at the register
Camera analytics can tell you the median seconds a guest spends from joining the line to placing their order. If you watch the curve over the day, you'll see your peak and you'll see how staffing decisions either soften or sharpen it. Most of our clients discover that the actual peak is 12–18 minutes earlier than they thought.
3. Dwell time at the menu
How long does the average guest stand in front of your menu before ordering? Under 30 seconds means the menu is doing its job. Over 70 seconds is a sign of either menu confusion or analysis paralysis — both fixable with menu re-engineering. The data isn't really about minutes; it's about whether your menu is helping or hurting throughput.
4. Queue abandonment
How many people joined the line and left before ordering? This is the silent killer for fast casual operators. If 9% of your weekend lunch arrivals abandon the line, you have an 18-second-faster-throughput problem and a 12% revenue gap, and you'd never see either on a P&L. Cameras see both.
5. Kitchen ticket-pickup time
From the moment the kitchen calls a ticket to the moment the runner picks it up. We've seen this go quietly from 14 seconds to 2 minutes during a dinner rush, with the obvious downstream effects on food quality and seat turn. The number is invisible without measurement and obvious once you see it on a chart.
6. Cash register events without staff present
This one is sensitive but real. Drawer-open events that don't pair with a transaction, when the camera shows a single staff member at the station with no customer present. If you have a leak (see our earlier post on POS leakage), this is where the camera tells you who and when, without you needing to spot it in the moment.
What this requires (and what it doesn't)
What it requires: cameras with reasonable angles on the door, register, kitchen pass, and dining room. A NVR or cloud video service with analytics enabled. A weekly half-hour to actually look at the data.
What it doesn't require: new hardware in most cases, an analytics specialist, a privacy-policy headache (none of these metrics involve identifying individual guests). The data is aggregate and operational.
Practical first step
Pull a single week of footage from your busiest two days, and answer one question: what's the gap between door-entries and tickets? If you're shocked by the number, you've found your highest-leverage operational metric for the next 90 days.