Video Surveillance & CCTV: Systems, Remote Monitoring & Cost (2026)

Cameras used to be a passive record you reviewed after something went wrong. Today, video surveillance is an active layer of security — cameras with AI analytics that flag intruders in real time, and remote operators who watch feeds live, warn intruders over a speaker, and dispatch police. Used well, video deters crime, provides evidence, and extends a small security team's reach across an entire property at a fraction of the cost of guards everywhere. Used badly, it's a wall of monitors nobody watches and blurry footage police can't use. This guide covers camera types and where each fits, the technology (resolution, IP vs. analog, storage and retention), AI analytics and their limits, the critical difference between recorded and monitored video, cybersecurity and NDAA-banned brands, the privacy law that trips up most employers, and what it all costs — with worked storage numbers, real buildout budgets, and the exact questions to ask a provider.

Camera types — and where each one fits

The single most common video mistake is buying the wrong camera for the location. Each form factor exists for a reason. Match the camera to the job — the field of view, the lighting, the distance, and what you actually need to identify — and a modest system outperforms a wall of the wrong hardware.

A key nuance: a PTZ can only look one direction at a time, so an unattended PTZ often misses the incident it was pointed away from — it shines when an operator or auto-tracking drives it. Thermal and LPR are specialist tools; don't expect a general dome to read plates at a gate or detect a person across a dark field. Most real sites use a mix: fixed domes/turrets for identification at choke points, bullets or thermal on the perimeter, and dedicated LPR at vehicle gates.

Resolution, lens and field of view — the basics that decide image quality

"We have cameras" means nothing if the footage can't identify a face or a plate. Three specs drive that:

• Resolution. 2MP (1080p) is the practical floor; 4MP and 4K (8MP) are common now. More megapixels help only if the lens and light support them — a 4K camera in the dark still gives you a smear.
• Lens and field of view. A wide lens covers more area but shrinks every subject; a narrow/telephoto lens covers less but captures detail at distance. You cannot get identification-grade detail across a huge area from one wide camera — that's the "blurry figure, no face" trap. Varifocal lenses let the installer dial in the exact field of view on site.
• Pixels on target. The professional way to spec cameras is pixels-per-foot (PPF) at the point that matters — not megapixels in the abstract. Design around faces at entrances and plates at gates, then fill in general coverage — not the other way around.

The DORI table: how many pixels you actually need

The industry standard for "how much detail" is DORI — Detect, Observe, Recognize, Identify — codified in IEC/EN 62676-4 and echoed by the U.S. Army's DCRI criteria. It maps a task to a required pixel density on the subject. Convert those to pixels-per-foot and it becomes a spec tool you can actually apply:

Turn that into distance. At identification grade (~80 PPF) with a matched lens, a 2MP camera holds detail to roughly 25 ft of scene width, a 4MP to roughly 40 ft, and a 4K to roughly 80 ft — and only across that width. Push the same camera to cover a wider area and PPF collapses: you drop from "identify" to "there's a person" fast. This is precisely why one wide camera can't identify faces across a big lot, and why pros place a tighter camera on the door and let a wide one handle general awareness. Also weigh low-light and infrared performance (most incidents happen at night), wide dynamic range for bright/backlit doorways, and frame rate. Overzoomed, wide-angle, or poorly-lit cameras are the top reason systems fail when it counts.

IP vs. analog vs. HD-over-coax

• Analog (legacy). Older standard-definition coax cameras to a DVR — declining, but still installed at many sites. Low resolution and no smart features.
• HD-over-coax (HD-TVI/CVI/AHD). High-definition analog that reuses existing coax cabling to a DVR — a budget upgrade path that avoids rewiring, but caps out on resolution and analytics compared to IP.
• IP / networked. Digital cameras over your data network to an NVR or VMS, usually powered by a single PoE (Power over Ethernet) cable. The dominant architecture today: higher resolution, built-in analytics, and easy remote access. Because they're on the network, they also carry cyber risk (see below).

Most new commercial installs are IP. HD-over-coax is a reasonable interim move when you're keeping existing wiring; new builds should go IP unless there's a specific reason not to.

Where footage lives: DVR vs. NVR vs. VMS vs. cloud/hybrid

• DVR records analog/HD-over-coax cameras to a local box with hard drives.
• NVR records IP cameras locally — higher resolution, more channels, network access.
• VMS (video management software) is server- or software-based management for larger, multi-site deployments — more cameras, richer analytics, user permissions, and integrations than an appliance NVR.
• Cloud / VSaaS hosts footage and management off-site: nothing on premises to fail or be stolen, updates and cybersecurity handled by the vendor, and access from anywhere — at a recurring per-camera cost and dependent on your upload bandwidth.
• Hybrid keeps a local recorder for full-resolution footage and bandwidth savings while mirroring clips or events to the cloud for offsite backup and remote access — increasingly the default for serious sites.

Retention math — a worked example, not just a formula

Retention is a budget decision, and the arithmetic is simple: storage ≈ cameras × per-camera bitrate × recording hours per day × retention days. But a formula doesn't answer the real question — how big a drive or cloud plan do I actually need? So let's work it. Take a common setup: 8× 4MP cameras using H.265 compression. A 4MP H.265 stream averages roughly 2 Mbps, which is about 20 GB per camera per day recording continuously — so 8 cameras generate about 160 GB/day. Switch to motion-/event-only recording on a low-traffic site and that drops to roughly a third:

Read it as a shopping list: for 90-day continuous retention on 8 cameras you're buying about 14 TB of usable space — realistically two 12 TB surveillance-grade drives in RAID for redundancy, a few hundred dollars once. The same footage in the cloud is a recurring few hundred dollars a month, but with offsite protection and no drive to fail. Two levers cut every number above: H.265/smart codecs roughly halve bitrate versus older H.264, and motion-based recording cuts storage several-fold on quiet cameras. The point: decide the retention you need first, then size the drives or cloud plan to hit it — don't let whatever the box shipped with decide how long your evidence survives.

AI and video analytics — what it does, and what it can't

Analytics are what turned cameras from recorders into an active security layer. The mature, widely-deployed capabilities:

• Person and vehicle detection — classify what triggered motion, so a swaying tree or a raccoon doesn't page anyone.
• Line-crossing and intrusion zones — alert when someone crosses a defined boundary (a fence line, a dock after hours).
• Loitering / dwell — flag someone lingering at an ATM, a doorway, or a loading area.
• Object left/removed, crowd/occupancy, and LPR — abandoned-package alerts, people-counting, and plate reads.

The biggest practical payoff is false-alarm reduction: filtering motion down to real human/vehicle events makes 24/7 live monitoring affordable, because operators aren't drowning in nuisance alerts. But know the limits. Analytics still generate false positives and misses — heavy rain, fog, glare, deep shadow, camera shake, and bad angles all degrade accuracy. "Facial recognition" is not the same as "person detection," is far less reliable in the wild, and is restricted or banned in some states and cities — treat it as a legal question, not just a feature. And analytics only alert; they don't respond. An alert with nobody watching is just a faster way to record a crime. Analytics amplify a monitoring plan — they don't replace one.

Recorded vs. live monitoring vs. video-verified response

This is the decision that most affects both security outcomes and cost:

The step change is active monitoring with talk-down: a trained operator sees the intruder, calls out a specific warning ("You in the gray hoodie by the blue truck — we see you, police are en route"), and most would-be intruders leave before anything happens. Because the operator has visually verified a real crime, the police dispatch is prioritized rather than treated as one more likely-false alarm — a decisive advantage that supercharges your alarm response. Many sites combine models: AI-monitored, talk-down-equipped cameras for after-hours active deterrence, with all footage recorded for evidence. Confirm the monitoring center is US-based, staffed 24/7, and ideally UL-listed / certified.

Remote monitoring vs. a guard post — the ROI, in dollars

The site's own headline is that monitored cameras cost "60–85% less" than guards. Here's the math behind it. A single 24/7 physical guard post is never one person — it takes roughly 4+ officers on rotation to cover 168 hours a week, which is why it lands at $16,000–$25,600/month unarmed (more for armed). A monitored, AI-driven multi-camera system covering the same property runs a fraction of that:

That's the 60–85% figure made concrete — and it's usually an unfair comparison in the cameras' favor, because a single guard post watches one spot while monitored cameras watch the entire perimeter at once. The catch: cameras can't lay hands on anyone. The strongest programs don't pick one — they lead with monitored cameras and add mobile patrol or a guard only where a body is genuinely required (see the decision matrix below).

Integration and camera placement — design coverage first, buy hardware second

A camera system is worth far more when it's tied to the rest of your security and placed with intent. Start with a coverage plan, not a hardware list: cover every entrance and exit (the choke points every incident passes through), parking and the perimeter, loading docks and cash-handling areas, and sensitive interior zones like server rooms and stockrooms. Overlap fields of view at critical points, position cameras to capture faces and plates (not the tops of heads), and mind lighting and glare. Then integrate:

• Access control. Every badge event gets a matching visual record — "a door opened at 2 a.m." becomes "here's exactly who opened it," and you catch tailgating and badge-sharing.
• Alarm monitoring. Cameras tied to alarm monitoring enable video verification, so an alarm becomes a confirmed, priority police call instead of a maybe.
• Guards and patrol. Cameras paired with guards or mobile patrol multiply both: monitored cameras cover the whole property cheaply and direct a smaller physical team exactly where they're needed, while guards provide the hands-on intervention cameras can't.

A set of disconnected cameras recording to a box in the closet is far less valuable than a smaller, well-placed system that's monitored and integrated.

Camera cybersecurity — and NDAA-banned brands

Every IP camera is a networked computer, and an unsecured one is a door into your network — a foothold for ransomware and a target for botnets. The basics are non-negotiable: change every default password (the single most exploited weakness in cameras), keep firmware patched, put cameras on a segmented VLAN separate from business systems, disable ports and cloud features you don't use, and avoid exposing recorders directly to the open internet.

There's also a brand and compliance dimension unique to the US. Under Section 889 of the FY2019 NDAA and the FCC's Covered List, video surveillance equipment from certain Chinese manufacturers — notably Hikvision and Dahua — is barred from federal contracts and federally-funded purchases. The article-warning below has the buyer's shortlist. The trap is OEM rebranding: this hardware is sold under dozens of other names, so a "not-Hikvision" label on the box means nothing. Ask your vendor to confirm the underlying manufacturer in writing, and check against a current list:

2026 update: This isn't just a procurement footnote anymore. In late October 2025 the FCC adopted rules ("Operation Clean Carts") creating a process to revoke or limit already-granted equipment authorizations for covered brands — not just block new ones — and major retailers pulled large numbers of Hikvision/Dahua listings in the ensuing delisting wave. Federal universal-service-funded entities face a Priority-1 removal deadline of May 8, 2026. Purely private buyers with no federal, grant, or lease nexus can still legally use these cameras, but the supply, support, and insurance picture is deteriorating fast — one more reason to spec NDAA-compliant from the start.

Privacy, notice and the law

Beyond audio consent, a few rules govern where and how you can watch:

• Reasonable expectation of privacy. Cameras in open, public-facing areas (sales floors, lots, entrances, hallways) are broadly permissible. Cameras in restrooms, locker rooms, changing areas, or similar private spaces are illegal in essentially every state — full stop.
• Employee notice. Several states require notifying employees of workplace monitoring, and some require it in writing; watching workers via camera can also intersect with labor law (e.g., you generally can't surveil protected organizing activity). Post clear signage — it supports notice, strengthens deterrence, and can matter legally.
• Retention, subpoenas and access. Your footage can be subpoenaed as evidence, and a written retention policy plus a documented chain-of-custody makes it usable and defensible. Control who can view and export footage, and log access.

The safe default for most private businesses: video-only in non-private areas, clear signage, written employee notice where required, and a documented retention and access policy.

What video surveillance costs — with worked buildouts

Cost depends on the number and type of cameras, storage, and whether you add live monitoring. Per-unit figures first, then three real systems built from them:

The biggest upfront driver is usually labor, not cameras — cable runs, mounting on masonry or high ceilings, trenching for gate/perimeter cameras, and network/PoE work — which is why "installed" cost (not the camera's sticker price) is the number that matters. Ongoing costs are storage and (if you choose it) monitoring, plus maintenance: firmware updates, lens cleaning, re-aiming bumped cameras, replacing failed drives, and periodic health checks. Because remote monitoring runs 60–85% less than equivalent guard coverage, many properties lead with monitored cameras and add patrol or a guard only where a physical presence is truly needed. See our cost guide for guard comparisons.

Cameras or guards? A decision matrix

They're complementary, not competing — but "it depends" is useless when you're writing the check. Cameras (especially monitored, with talk-down) deter, record, extend coverage cheaply, and add real-time response for a fraction of a guard's cost. What they cannot do is put a human body between a threat and your people or assets. Use this to decide where a camera is enough and where a physical guard is non-negotiable:

The strongest, most cost-effective programs pair a monitored camera system with a smaller guard team or mobile patrol rather than choosing one or the other — a pattern that also anchors retail loss-prevention programs, where cameras, analytics, and floor staff work together. If you're weighing armed vs. unarmed for the guard side, see armed vs. unarmed guards.

How long to keep footage

Retention clusters at 30, 60, or 90 days for most commercial sites — 30 days covers typical insurance and police timelines, and 60–90 days suits higher-risk operations. Some sectors keep footage far longer (casinos and bank vaults up to a year). A common myth is that PCI or HIPAA mandate a specific retention period for all cameras — they don't: PCI's 90-day figure applies to video or badge data for sensitive cardholder areas (data centers, server rooms), not the retail floor, and HIPAA sets no fixed video-retention period. The right approach is a written retention policy by camera class based on your risk and any sector-specific rules — then size storage to hit it (use the retention table above).

Why most camera systems fail when it matters

The dirty secret of video surveillance is that a large share of installed systems are useless at the exact moment they're needed. The failure modes are predictable: cameras aimed too wide, so footage shows a blurry figure but no identifiable face or plate; poor low-light performance, so the overnight incident you most care about is a smear; footage overwritten before anyone pulled it because retention was set to whatever the drive held; a camera knocked out of aim or offline for weeks with no one alerted; and — most common — nobody watching, so the system records the crime without doing anything to stop it. When police ask for footage, "we have cameras" too often becomes "the image is useless." Avoiding this is about deliberate choices, not more cameras: position and zoom cameras to hit identify-grade PPF on faces at entrances and plates at gates, buy real low-light or infrared capability where you need night coverage, set retention by policy rather than by disk, monitor camera health so outages get caught, and decide whether recorded-only is enough or the site warrants live or AI monitoring. A smaller system designed to actually identify people beats a wall of cameras that produce evidence no one can use.

How to choose a provider — the questions that matter

Decide first whether you need recorded, monitored, or AI-monitored video. Then get past the glossy checklist (US-based, 24/7, NDAA — all necessary but table stakes) and ask the high-leverage questions that separate a good provider from a lock-in trap:

• What's the monitoring-center SLA? Get the alarm-to-dispatch chain in writing: how fast an AI alert reaches a live operator, how fast they verify, and the target dispatch time. Confirm the center is US-based, staffed 24/7, UL-listed, and redundant.
• Who owns the cameras and the footage? This is the biggest lock-in trap. Insist that you own the hardware and recordings, that the system uses an open standard (ONVIF) rather than a proprietary-only recorder, and that footage is exportable in a standard format. A closed ecosystem can hold your own video hostage.
• What are the contract length and auto-renewal terms? Watch multi-year lock-ins with evergreen auto-renewal and steep early-termination fees — especially when "free" cameras are bundled to bury a long monitoring commitment.
• Are there price escalators? Many monitoring contracts include annual increase clauses. Ask for the number and cap it.
• What happens if I leave? Do the cameras keep working on your own recorder? Do you get every clip of your footage? If the answer is "the system stops" or "the footage stays with us," walk.
• Who is the true manufacturer? Get NDAA compliance and the underlying maker in writing, given the rebranding problem above.

Red flags: no written monitoring SLA; provider-owned footage you can't export; proprietary-only (non-ONVIF) hardware; auto-renew beyond one year with a large early-termination fee; a vendor that won't name the camera's actual manufacturer; and "free" equipment that only makes sense once you read the multi-year monitoring contract. Confirm licensing where the provider also supplies guard or monitoring services — the same diligence you'd apply when you hire a guard company. Reputable providers on our network cover Los Angeles, Chicago, and cities nationwide.

Ready to design a system? Get free quotes from licensed security companies that offer video surveillance, or explore video surveillance and CCTV services in your area.