Most camera projects don't fail because of the cameras. They fail because of decisions made before a single camera is mounted: the objective nobody wrote down, the light nobody tested for, the storage nobody sized backward from a retention number. I design camera systems for a living, and when I get called in to fix a deployment, it is almost always one of the same five mistakes. Here is each one, why it keeps happening, and the design call that prevents it.
Mistake 1: Designing for coverage instead of identification
It is tempting to put up one wide camera, see the whole area on the monitor, and call it covered. But coverage and identification are not the same thing. The industry measures the difference in pixels on the target, and there is more than one objective. Each one needs dramatically more pixel density than the last.
| Objective | Pixel density on target | What it answers |
|---|---|---|
| Detection | about 25 px/m (8 px/ft) | Something is there |
| Recognition | about 125 px/m (38 px/ft) | Is that someone you know? |
| Identification | about 250 px/m (76 px/ft) | Name a stranger |
The most common version of this mistake is one high-resolution camera over an entire parking lot. On the monitor it looks like full coverage. Six months later there is an incident, the footage exists, and there is not enough detail on the target to identify anyone. The camera worked exactly as designed. The design didn't. The rule: state the objective before you choose the model. If you cannot name the job a camera has to do (detect, recognize, or identify) you are not ready to select it.
Mistake 2: Treating low light as an afterthought
I have never once been called to a site because a camera looked bad at noon. Every emergency call happens after dark, in rain, glare, headlights, or against a backlit doorway. Two things decide whether a camera delivers when it matters: how it handles strong backlight (wide dynamic range) and how it really sees in near-darkness (true low-light sensitivity and IR). Spec these for the hour the incident will actually happen, and walk the site at night, not at the convenient mid-day visit. If a camera cannot deliver a usable image at a backlit entrance after dark, nothing on the spec sheet will save it.
Mistake 3: Making power and cabling an afterthought
Cameras run on Power over Ethernet, and PoE has hard limits. A copper run tops out around 100 meters, about 330 feet, and every camera draws from the switch's power budget. A heated housing or a pan-tilt-zoom pulls far more than a small fixed dome, and it adds up fast across a 24-port switch. Here is how it plays out on a real site: a few cameras won't boot or keep dropping link, the camera gets blamed, the switch gets swapped, and the actual problem was power budget, cable distance, or a bad termination the whole time. Add up the PoE load and verify the run lengths before you buy the switch, not after the install, when half the cameras are dark.
Mistake 4: Guessing at storage instead of designing it
This is the one that hurts months later. Before you buy a single camera, answer one question: how long do you need footage available when someone asks for it? Thirty days? Ninety? A year for compliance? Storage is designed backward from that retention number, and after that it is just resolution times frame rate times camera count times days. The mistake is buying beautiful 4K cameras, keeping a week of footage, and discovering, the day you need video from three weeks ago, that it was overwritten long before. Pick the retention window first; the recorder and the drives follow from it.
Mistake 5: Expecting one camera to do everything
A pan-tilt-zoom looks like it can cover everything, until you remember it only records where it is pointed. It is the one camera in your system that can be looking away from the incident. That is not a flaw; it is exactly why experienced designs pair tools instead of forcing one to do every job.
| Fixed | PTZ | Multisensor | |
|---|---|---|---|
| Whole view recorded at once | Yes | No | Yes |
| Identification detail at distance | No | Yes | Partial |
| Follows / investigates | No | Yes | No |
| Moving parts | None | Motors | None |
| Best for | Chokepoints, critical doors | Large open areas you actively watch | Wide areas needing constant detail |
A fixed camera holds the critical door it never leaves. A multisensor covers a wide area at constant detail with no moving parts. A PTZ gives an operator the reach to investigate at a distance. Match the form factor to the job, and budget for the wear and upkeep a PTZ's motors add over a fixed or multisensor camera.
The thread through all five
Underneath every one of these is a single principle: every camera position should have a job. Identify faces at the entrance. Read plates at the gate. Watch the dock during deliveries. Verify who badged through that door. If a position does not have a clearly defined job, a camera probably should not be there, and if it does, that job tells you the resolution, the lens, the low-light spec, and the storage it needs. The skill was never picking a camera. It is designing an architecture where each tool does what it does best.
The takeaway: State the objective before you choose the model. Detect, recognize, or identify, every camera position needs a job, and the job sizes the camera, the lens, the light it has to beat, and the storage behind it.
The one thing to check this week
Pick one camera in your existing system, ideally one watching a parking lot or a wide outdoor area, and ask the people who own it what its job is. If the answer is "general coverage," pull up actual recorded footage from that camera and try to identify a face or a plate at the far edge of the frame. If you can't, you have just found a position that is detecting but not identifying, and you now know exactly what to change the next time that area gets re-specced.
If you want a second set of eyes on a camera design, or help matching the right fixed, PTZ, or multisensor to each position, our senior specialists do this every day. Browse cameras from Axis, Hanwha, i-PRO, and more in our IP camera catalog, or call 877-277-7147 to talk it through.