The question is what each camera has to prove
Walk into most businesses with a camera system and you will find the same thing. Sixteen cameras, all recording, none of them useful. The one covering the register is mounted too high and catches the top of everyone’s head. The parking lot camera is a wide shot where every car is four pixels. The entry camera faces a glass door, so every daytime image is a silhouette against blown out sunlight.
None of that is a hardware problem. It is a specification problem, and it happens because the system was priced by camera count instead of designed around what it needs to capture.
We start the other way around. For each position we ask what it has to prove. Identify a specific person at the door. Read a plate at the gate. Establish that a pallet left the dock at a particular time. Each of those implies a different resolution, a different lens, a different mounting height and a different tolerance for backlight. Get that right and you need fewer cameras, not more.
Design
Site survey first. We walk the property and plan field of view per position, accounting for what the light actually does at different times of day. West facing entries are a known problem. So are loading doors where the camera is looking from dark into bright.
Resolution matched to distance. Pixels per foot at the target, not megapixels on a datasheet. A 4K camera pointed at a large yard can easily deliver worse identification than a well placed 2MP camera at the right distance.
Low light handled deliberately. Most incidents happen when nobody is around, which is also when the lighting is worst. Infrared, low light sensors and supplementary lighting each solve different parts of that, and infrared reflecting off a nearby wall or falling snow will wash out an image entirely if the placement is careless.
Coverage that overlaps where it matters. Entries and exits get redundancy. General areas do not need it.
Recording and retention
Retention is a business decision before it is a technical one. Thirty days covers most situations and matches what insurers generally expect. Some contracts and some regulated environments require considerably more.
We size storage to the retention you actually need, at the frame rate and resolution you are actually recording, rather than quoting a recorder and hoping. Continuous recording, motion triggered recording and a hybrid of the two produce very different storage requirements, and the hybrid is usually right: continuous at low frame rate with motion triggering a higher rate.
Cloud recording is worth considering where a break in could remove the recorder itself, or where you have many small sites and no appetite for hardware at each one. It costs more monthly and depends on upload bandwidth, so it fits some situations and not others. We will tell you which one you are.
The security of your security system
This is the part that gets overlooked, and it is genuinely serious. Camera recorders are among the most commonly compromised devices on business networks. They ship with default credentials, they are frequently exposed directly to the internet so somebody can view from a phone, and they are rarely patched.
A compromised recorder is bad twice: an attacker sees your premises, and they have a foothold inside your network from a device nobody is monitoring.
We put cameras and recorders on a segmented VLAN that cannot reach your servers or workstations, change every default credential, keep firmware current, and provide remote viewing through properly secured access rather than a forwarded port. This is the same approach described in our network security work, applied to the devices that most often get left out of it.
Integration with access control
Cameras and doors on separate systems means an incident review involves pulling a badge log from one application, working out a timestamp, then scrubbing footage in another.
On an integrated platform the badge event links to the footage of that moment. Someone used a credential at the back door at 11:40pm, and the clip is one click away. If you are deploying both, doing them together on one platform is significantly more useful than buying them separately. See access control for that side of it.
Installation
Cameras are low voltage work and the cabling determines whether the system stays reliable. We run it as part of our structured cabling practice: proper pathways, correct cable for the run length, weatherproof terminations outdoors, and Power over Ethernet budgeted so the switch can actually carry the load with heaters running in winter.
Mounts are specified for the surface and the exposure. Every camera is aimed, focused and commissioned rather than pointed roughly and left. You get documentation showing what each camera covers, which is the thing nobody has when they need to find footage in a hurry.