How Many Cameras Does a Factory Floor Actually Need?

A factory floor needs about one camera per question you need answered on it — not a count based on floor area. In practice that means one camera per active work zone, one per entry/exit, and one per aisle crossing. A 1,500 m² shop with 6–8 machine cells, two doors and a materials aisle usually lands around 8–12 cameras.

Treat that number as a sanity-check, not a formula. Two floors of identical size can need very different counts, because they need different things seen.

Key takeaways

• Count by the questions you need answered, not by square metres or a per-m² ratio.
• An operations count (is this cell running?) is usually lower than a security count (who took that pallet?) — decide which you are buying before you count.
• One wide 2.8 mm lens everywhere quietly fails the read-a-label / read-a-face job — that is the DORI trap.
• Obstructions add cameras: a tall machine in a corner hides a wedge that grows with distance.
• Coverage = placement × availability. A camera that is dark during a power cut covered nothing.

Start with the questions, not the rooms

A camera count is a demand-side number. Before you size hardware, write down what you actually need to see:

• Can I tell if machine cell 3 is running right now, or idle?
• Can I read who signed for that pallet at the dock?
• Can I tell why the line stopped at 14:00?
• Can I see whether the forklift aisle was clear when someone crossed it?

Each of those is a camera — or, for the "read a detail" questions, a specific camera on a specific sightline. A camera doesn't cover an area; it answers a question about a spot. That is why square footage tells you almost nothing. Sightlines do.

So the planning unit is the zone: any spot where work, movement, or risk you need to see actually happens — a machine cell, a doorway, a forklift crossing, a hazard.

Add them up. That is your starting number — before you adjust for the room.

Are you counting to secure the floor or to run it?

Almost every camera guide online silently assumes you are buying security: theft, perimeter, dock LPR, faces at the gate. That count is high, because security cameras have to identify — read a plate, a face, a badge.

An operations camera has an easier job. To know that cell 3 is cutting, that the line is flowing, that a zone went idle at 14:00, the camera only has to see activity — not identity. A single wide view of a zone answers "is this running?" without ever resolving a face. So the operations count is usually lower than the security count for the same floor, and the cameras sit differently: over the work, not just at the doors.

Decide which floor you are building before you count. Most Indian factory owners we talk to want both eventually, but they are buying operations first — uptime, flow, bottlenecks — and paying for a security-grade count they don't yet need.

The lens trap most floors fall into

Here is the failure you don't discover until you need the footage: one 2.8 mm wide lens on every camera.

A 2.8 mm lens sees roughly a 90–110° cone — excellent situational awareness, a whole zone in one frame. But wide-angle spreads your pixels thin. The governing standard, IEC 62676-4 (DORI), sets a minimum pixel density on the target for each job:

A factory camera does one of two jobs: situational awareness (is the zone active?) or identification (read this detail). They need different lenses. A 2 MP wide camera easily clears Detect/Observe across a big zone, but it drops below Recognise/Identify within a few metres — so the same camera that shows the whole cell can't read the label, gauge, or badge across it. One wide lens everywhere is a blind spot hiding in your pixel budget, not on your floor plan. "Placement over resolution" really means: put the required pixels-per-metre on the zone that needs them.

Then correct for the room

Zones give you the count. The building takes some back.

• Obstructions add cameras. Tall racking, mezzanines and large machines block sightlines. A column or machine of width w at distance d hides a wedge that only grows with range — so a cluttered floor needs more cameras than an open one of the same size. Budget a buffer for it.
• Mount to the view you need. Bullet cameras are usually wall-mounted and often paired with longer lenses for a narrow view down an aisle; turrets install easily on a ceiling and are typically used for wider top-down coverage. The view is a lens choice, not a property of the housing — but the conventions hold on most floors.

• Overlap lightly, on purpose. A small overlap at zone edges removes the seams where an event slips between two views. Aim for light overlap, not full redundancy.
• Don't over-cover dead space. Empty walkways and unused storage don't each need a camera. Cover where the value and the risk sit. Leave the rest.

The India layer nobody counts

On an Indian floor, a few drivers change where the cameras go — and how many:

• People and contractors, not just theft. Labour- and contractor-heavy floors monitor density and activity across shifts, which pushes coverage over the work area, not just the perimeter.
• Safety obligations. Factories Act duties around hazardous machinery and restricted zones make certain areas mandatory-cover, regardless of their floor area.
• The April 2026 rule. From 1 April 2026, CCTV sold in India must carry STQC certification (Essential Requirements under MeitY) plus BIS registration under the Compulsory Registration Order, and Chinese-origin OEMs are not eligible. That shrinks the certified hardware list and raises per-unit cost — which makes an efficient, question-driven count (fewer, better-placed cameras) worth more than blanketing the floor.

Field notes from a live floor

We run a working pilot on a metalworking and engine-repair shop. Two things stood out that no area formula predicts.

In Mama's Mihnovka pilot (2026), 3× 2 MP / 2.8 mm cameras covered a live production floor for 14 days and logged exactly 1 power outage — daytime, 0 at night, 0 reboots. The laptop kept recording on its battery; the PoE switch, router and all three cameras had no battery and dropped instantly for the ~1h50m the mains was out. On an Indian floor that isn't an edge case. Coverage isn't just placement — it's placement × availability, and a camera that's dark during the one daytime outage covered zero zones on that shift. Put a small UPS on the PoE switch and router. It matters as much as the camera count. (Honest caveat: this is one outage over 14 days — indicative, not a base rate.)

Placement was decided by walking, not by a grid. A tall vertical milling machine stands in one corner of the floor — exactly the kind of obstruction that throws a shadow a plan-view grid never accounts for. You catch it by standing behind it, not by drawing squares on a layout. The count is the easy part. Where the cameras go — and what keeps them powered — is what decides whether you see what you need to.

The fast way to get the number right

Counting zones by hand is slow and easy to botch on a busy floor. That's the job we built Mama for: film a two-minute walkthrough on your phone, and it returns a count, a placement camera-by-camera (ceiling or wall), and a coverage map — no engineer, no site visit, same day.

FAQ

How many CCTV cameras do I need for a small factory? A small factory typically needs 6–12 cameras. Count one per active work zone, one per entry/exit, and one per aisle crossing, then adjust for obstructions. The number follows your layout and what you need to see — not floor area.

Should I count cameras for security or for operations? Decide first. A security count has to identify faces and plates, so it runs high. An operations count only has to see activity at a zone, so it's usually lower — and the cameras sit over the work, not just at the doors.

Is it better to have more cameras or higher-resolution cameras? Placement beats resolution. A 2 MP camera on the right sightline sees more that matters than a 4K one on a blind wall. Get coverage right first, then raise pixel density where you need to read detail — labels, faces, gauges — per the DORI standard.

Ceiling or wall mounting for factory cameras? Use both. Ceiling turrets suit wide top-down coverage of a work zone; wall bullets suit long, narrow views down an aisle. The view comes from the lens, not the housing, but those pairings hold on most floors.

Do I need a camera in every area of the plant? No. Cover where value and risk sit — work zones, exits, hazards, high-value storage. Empty walkways and unused space don't each need a camera.