Camera systems for businesses that will have to produce the footage
A business camera system gets judged once: the day an insurer, a detective or an employment lawyer asks for a clip. We design for that day: identification-grade coverage where the money and the doors are, verified hardware, and a retention figure that holds up to arithmetic.
Three audiences you are actually filming for
A residential camera answers a question for the person who owns it. A commercial camera answers a question for somebody else, usually under pressure and usually months later. There are three of them and they want different things.
Your insurer wants continuous coverage of the incident with a defensible timestamp, and will discount or deny a claim where the only footage is a 3 fps clip that skips the two seconds that matter. A detective wants a face or a plate at a density that can be put in front of a witness, exported in a format that opens on a computer they control, without a proprietary player they have to install. An employment attorney, yours or the other side’s, wants to know exactly where every camera was pointed, what the retention policy said, and whether anybody was recorded somewhere the law says they should not have been.
Designing for those three tends to produce a different system than designing for a demo on a wall-mounted TV. It means identification-grade pixel density at the doors, the register and the dock rather than even coverage everywhere. It means export that produces a standard file plus a log entry, not a screen recording of a phone app. It means the retention number is arithmetic and not a marketing figure. And it means a written policy about who can view and who can export, agreed before the first camera goes up.
If you want the underlying design math (pixel density bands, lens selection, storage formulas) it is all laid out on our security camera installation page. This page is about the parts that only apply when the property is a business.
What earns its place, by type of business
| Business | The cameras that earn their keep | The one that is usually missing |
|---|---|---|
| Retail, bodega, pharmacy | A tight camera over the register facing the customer side. One at every customer door at face height, not ceiling height. One covering the stock room door and one on the high-value fixture. | A camera on the drawer itself. Most internal shrink happens at the drawer, and a dome twelve feet up in the ceiling grid cannot show whose hand was in it. |
| Restaurant and bar | Point of sale, the service bar, the walk-in door, the delivery door and the alley behind it, and the route between the safe and the office. | The delivery door. Walk-ins empty at six in the morning through a propped fire door, not at two a.m. through the front. |
| Warehouse and distribution | Every dock position at plate-and-face density, the pick line, the returns bench, the yard gate, and high-bay overview for sequence. | A camera aimed along the trailer rather than across the dock. Across the dock you get a truck. Along it you can count what came off. |
| Office and professional | Suite entry, reception, the IDF or server room door, and any exterior door on a reader so the credential event and the face line up. | The IDF door. When a laptop, a switch or a drive walks, that is the only camera anyone asks for, and it is almost never there. |
| Parking, garage, yard | A dedicated plate camera as close to head-on as the geometry allows, plus a separate wide camera for vehicle color, make and occupancy. | The realization that one camera cannot do plates and context at the same time. Plate work needs a fast shutter and a narrow view; context needs the opposite. |
The pattern underneath all five rows is the same: a small number of tight cameras that produce identifiable people at the points where money, goods or access change hands, plus the minimum number of wide cameras needed to stitch those moments into a sequence. Typical, subject to survey. The geometry of your space decides the count.
NDAA Section 889 and the FCC Covered List: who this actually catches
Section 889 of the FY2019 National Defense Authorization Act names five companies and their subsidiaries and affiliates: Huawei, ZTE, Hytera, Hangzhou Hikvision Digital Technology and Dahua Technology. It has two prohibitions and the second one is the one that surprises people.
Part A, effective 13 August 2019, stops federal agencies buying covered equipment, or any system that uses it as a substantial or essential component. Part B, effective 13 August 2020, stops federal agencies contracting with any entity that uses such equipment anywhere in its own operations, not just in the federal deliverable. It is implemented through FAR 52.204-25, which is included by reference in essentially every federal contract above the micro-purchase threshold and requires an affirmative representation.
Read that again in plain terms: a company can lose its ability to bid federal work because of a camera watching its own back parking lot. The exposure attaches to the organization, not to the project.
Who is caught, and who genuinely is not
Caught: federal prime contractors and their subcontractors, GSA schedule holders, and (the one that matters most in New York) recipients of federal grant money, because the restriction flows down through federal award terms. That sweeps in a very large population of local nonprofits, houses of worship and community organizations funded through DHS Nonprofit Security Grant Program awards, most of whom have no idea that the camera package a well-meaning volunteer bought online in 2019 is now a compliance problem attached to their next application.
Not caught: a purely private commercial buyer with no federal contract and no federal grant money, and a purchase made entirely with state or local funds. If that is you, nobody is going to make you replace anything. Section 889 is a procurement rule, not a general ban.
The rebrand trap
The single biggest source of accidental non-compliance is that covered manufacturers reach the US market under other labels. Several familiar consumer and value brands have historically been built on covered-manufacturer platforms, corporate ownership of some of them has changed since, and a second and subtler exposure exists where a camera sold under an entirely unrelated brand runs a system-on-chip from a covered manufacturer’s silicon subsidiary. The box at receiving frequently says something different from the FCC ID inside it.
So the verification is per device, not per brand: capture make, model and FCC ID for every unit, cross-check against the FCC Covered List as it currently stands, and obtain the manufacturer’s compliance statement in writing, naming FAR 52.204-25 and signed by an officer. A number of well-known camera and video management manufacturers publish exactly that statement; the statement is the deliverable, not the reputation. We will not put a blanket compliance claim in a proposal. We will put the model list, the FCC IDs and the manufacturer statements in it, because on a federally funded job a loose compliance claim is a false-claims problem rather than a marketing problem.
Separately, and worth keeping distinct: the FCC Covered List, the equipment-authorization restrictions that followed the Secure Equipment Act, and Commerce export controls are three different regimes with three different scopes, and the practical scope has been moving. We verify current status against the FCC’s own Covered List page at the time of the bid rather than repeating what was true last year.
The cloud math nobody puts on the proposal
Cloud video is genuinely good at three things: there is no recorder on site for anyone to steal, firmware is managed for you, and remote access works without any inbound firewall rules. Those are real advantages and for some businesses they settle it. What gets left out of the pitch is bandwidth and five-year cost.
Continuous cloud recording needs roughly 1 to 2 Mbps of sustained upload per 1080p camera, and some platforms specify a committed 4 Mbps per camera. Twenty cameras is therefore 20 to 40 Mbps of upload that never stops; forty cameras is around 80 Mbps. Most commercial internet circuits in New York City are asymmetric and cannot carry that without a fiber upgrade, and if it is not shaped with quality-of-service it will make your phone system unusable and your card terminals slow. That conversation belongs before the order, not after the first Monday lunch rush.
Then the money. US cloud video runs roughly $3 to $60 per camera per month; most business plans land between $10 and $30 for thirty-day retention, with 4K plus long retention plus analytics reaching the top of the range. Forty cameras at $20 a month is about $48,000 over five years, before any circuit upgrade. Compare that against an on-premises recorder with essentially no recurring fee but a drive replacement cycle every three to five years at $100 to $200 a drive. Both of those are typical figures, subject to your actual site and camera mix.
The recommendation we give most commercial clients is hybrid, and it is not a compromise for its own sake: record at full rate locally where bandwidth is free, push events, clips and a low-rate sub-stream to the cloud for off-site redundancy and remote viewing, and keep the ability to pull any archive segment on demand. You get the theft-and-fire protection that a local recorder alone does not give you, without paying to upload eight hours a day of an empty stockroom. Where a business has multiple sites and no IT staff, full cloud often is the right call, but it should be chosen with the upload figure and the sixty-month number in front of you.
What AI detection actually does, and the four places it still fails
Object classification is the biggest practical improvement in commercial surveillance in a decade. It is also routinely oversold, so here is the honest version.
- What it replaces. Old motion detection compares pixels, so rain, snow, a headlight sweep, a tree shadow, a plastic bag and a spider on the lens all count as motion. Current edge analytics classify the object as a person, a vehicle or neither before raising an event, and in practice that removes the overwhelming majority of nuisance alerts, which is what makes event notification usable rather than something you mute in week two.
- What you get beyond alerts. Line crossing with direction, region entry and exit with dwell time, loitering, object left behind or removed, people counting, queue length and heat mapping. For a retailer the counting and heat data is frequently more valuable day to day than the security function.
- Failure one: the target is too small. Every analytics engine has a minimum object size in pixels, and a camera that satisfies your eye at 60 feet may be well under it. We draw the detection line inside the reliable pixel-density band rather than at the far edge of the frame, and we tune minimum and maximum object size rather than leaving the defaults.
- Failure two: the angle is wrong. Classification wants a near-horizontal view of a human form. A camera mounted flat to the ceiling looking straight down sees a hat and a pair of shoulders, and performance falls off badly. Mounting height and tilt are analytics decisions, not aesthetic ones.
- Failure three: backlight and occlusion. A silhouette against a glass storefront and a person half hidden behind a shelf both defeat classification. This is the second reason true wide dynamic range gets specified on entrance cameras, the first being that you want a face rather than an outline.
- Failure four: the lens is dirty. Spider webs across an infrared beam, road film on a garage camera and hazed dome bubbles quietly degrade analytics long before anyone notices the picture looks worse. This is the single most common reason a system that worked in year one stops raising useful events in year three, and it is a maintenance item, not a hardware fault.
Employees, biometrics and audio: the three legal edges in a commercial install
Employees. NY Labor Law §203-c bars an employer from causing video recording of an employee in a restroom, locker room, or any room designated for employees to change clothes. Footage taken in violation cannot be used for any purpose, and the section carries a private right of action. GBL §395-b covers the same physical spaces from a different direction and treats each device as a separate violation, with a narrow fitting-room exception requiring conspicuous posted notice, which in New York City must be in both English and Spanish.
Biometrics. NYC Admin Code §22-1201 applies specifically to retail stores, places of entertainment, and food-and-drink establishments. Businesses in those three categories that collect or retain biometric identifier information must post clear and conspicuous notice at every customer entrance, and there is a separate absolute prohibition on selling, leasing, trading or otherwise profiting from that information. There is a private right of action. Face detection that merely counts or blurs is not the same as face matching that returns an identity, but because the obligation can attach to retention alone, the right sequence is to talk to your counsel before enabling any matching feature, not after. Our position is that we will tell you when a feature you are being sold creates an obligation you did not know about.
Audio. New York is a one-party consent state, but a camera microphone in a dining room or a corridor is not a party to anything. Recording a conversation between two other people with no participant consenting is mechanical overhearing by a person not present, and under Penal Law §250.05 that is a class E felony. Camera audio ships disabled on our installs unless there is a participating party or posted consent your attorney has approved.
Where the money goes on a commercial job
Installed cost in New York City commonly lands between $500 and $2,500 per camera all-in, and the labor share alone is typically $150 to $300 per camera. Whole-system ranges we see in this market: eight to sixteen cameras in commercial space, roughly $6,000 to $18,000; twenty-four cameras and up on a campus or a large warehouse, $20,000 and upward. Network infrastructure (PoE switches, a router capable of the segmentation, drops) commonly adds $500 to $3,000 or more. All typical, all subject to survey.
The spread has almost nothing to do with the camera. It is cable route through a pre-war partition versus an open warehouse ceiling, firestopping every rated penetration, switch capacity for cameras with heaters, storage sized to a real retention figure rather than to whatever drive came in the box, and whether the building limits when work can happen at all and requires a freight-elevator window.
What we hand over at the end is a camera schedule listing every position with its lens, its field of view and its design pixel density at the target distance; the retention calculation with the numbers used; the network diagram and VLAN configuration; a labeled cable map; and every credential registered to your business. If you replace us next year, your next contractor should be able to work on the system in an afternoon rather than reverse-engineering it for a week.
Common questions
How long should a business keep footage?
There is no general New York statute telling a private business how long to retain video, so the number should come from three places rather than from a default setting. First, whatever your insurer, franchisor or industry regulator requires in writing. Check, because some policies do specify it. Second, how long a claim typically takes to surface in your business: a slip-and-fall or an employment complaint routinely lands weeks or months after the event, while a police request on a theft usually arrives within two to six weeks.
Third, and least glamorous: what your storage will actually hold. Thirty days is the common commercial baseline and sixty to ninety days is normal where claims are slow to appear. If the budget will not carry uniform retention, tiered retention is the answer: keep entrances, registers and docks for ninety days and general areas for thirty, which typically cuts total storage by 30 to 40 percent. Whatever you land on, write it down as a policy, because the first question in litigation is not what the footage showed but why it was gone.
Can I put cameras where my employees work?
In the open work areas, generally yes. The hard prohibitions are specific: NY Labor Law §203-c bars video recording of employees in restrooms, locker rooms and designated changing rooms, and recordings made in violation cannot be used for any purpose. GBL §395-b covers the same spaces and makes each device a separate violation.
Beyond those lines the practical advice is about how you use it rather than whether you may install it. Tell staff the cameras exist and where; put the policy in the handbook; decide in advance who may view and who may export, and log it. Where a workplace is unionized or covered by a collective agreement there may be notice or bargaining obligations that have nothing to do with the camera code, and that is a question for your employment counsel before we mount anything. And audio stays off: a microphone in a break room recording two employees talking to each other is a felony exposure, not an HR one.
Does NDAA Section 889 apply to my business?
Ask one question: does any federal money reach your organization? A federal contract, a federal subcontract, a GSA schedule, or a federal grant, including a DHS Nonprofit Security Grant, which is extremely common among New York houses of worship, schools and community nonprofits. If the answer is yes, Part B means covered equipment anywhere in your operations is a problem, not only equipment in the federal deliverable.
If the answer is genuinely no (a private business spending private money) then Section 889 does not bind you and nobody is going to make you replace working cameras. That said, there is a separate and non-legal reason to plan a transition: following enforcement pressure through late 2025, major retailers removed enormous numbers of listings for restricted Chinese electronics, so replacement parts and firmware updates for that installed base are becoming difficult to obtain through compliant channels. Unpatched cameras with no support path are a security liability regardless of anyone’s procurement rules. We would plan a phased replacement over two or three budget cycles rather than a rushed rip-out. The existing cable and PoE infrastructure is usually reusable, which preserves a large share of what the original installation cost.
Is cloud video worth it for a small business?
Sometimes clearly yes: multiple small sites, no IT staff, nobody on the premises to notice a recorder has been stolen, and a manager who needs to look at four locations from a phone. In that shape, cloud earns its fee.
Run the two numbers first. Upload: roughly 1 to 2 Mbps sustained per 1080p camera, so eight cameras is 8 to 16 Mbps that never stops, and you need to know your circuit’s real upload speed rather than the number on the bill. Cost: at $10 to $30 per camera per month, eight cameras is $960 to $2,880 a year, every year. Compare that against a local recorder with a drive swap every three to five years.
For most single-site businesses in this market the answer we give is hybrid: full-rate local recording plus cloud for events, remote viewing and off-site copies of the clips that matter. It covers the genuine weakness of a local recorder, which is that a fire or a thief takes the evidence along with everything else, without paying monthly to upload an empty stockroom.
Can you connect cameras to our point-of-sale or access control system?
Yes, and on a retail or restaurant job it is usually the highest-value integration available. Point-of-sale text overlay writes the transaction line onto the video record, so a manager can search for voids, no-sales, refunds and manual price overrides and jump straight to the clip instead of scrubbing three hours of a register. It is the fastest way to close out an internal-loss question, and it takes a serial or network feed from the POS plus a recorder that supports the feature.
Access control integration is the same idea at the door: a credential event and a face on the same timeline, so you can see whether the person who badged in is the person the badge belongs to. That is the standard way tailgating gets identified and it makes an access audit meaningful. Both are design-time decisions: the recorder or video management platform has to support them and the network has to let the systems talk, which is one more reason to have the same contractor do the cameras and the access control.
How do we get footage to the police without handing over the recorder?
Decide this before you need it. The system should be able to export a defined time range as a standard file that opens on a computer you do not control, with the date and time burned in or carried in the file metadata, plus the manufacturer’s player as a fallback rather than as the only option. Proprietary-only export is where a lot of small systems fail: an officer with a USB drive and forty minutes will not install a vendor player on a department machine.
Set the process up as well as the technology. Name the two or three people authorized to export, keep a simple log of what was exported and to whom, and preserve the original on the recorder rather than deleting it once a copy leaves the building. If a matter is likely to become litigation, the retention clock needs to be stopped on that footage deliberately. Systems overwrite on schedule and will not wait for your attorney. We will show whoever holds the keys exactly how to do all of this at handover, because a system only one person can operate is a system that stops working the week they take vacation.
Work that usually comes with this
Security Camera Installation
New IP camera systems for buildings, storefronts, offices and homes.
Security Camera Repair & Upgrades
Systems that stopped recording, cameras that went dark, analogue-to-IP migration.
Small Business & Retail
Storefronts, bodegas, salons and offices: cameras, buzz-in entry and back-of-house control.
Restaurants & Bars
POS-overhead coverage, delivery doors, walk-ins, and staff-area access.
Warehouses & Industrial
Loading docks, yard cameras, gate control and high-bay coverage.
Structured Cabling
Cat6/6A and fibre backbones, racks, patch panels, certified and labelled.
Send us the floor plan and the loss you are trying to stop.
We’ll come out, walk the doors and the money, and come back with a camera schedule that names the lens, the distance and the pixel density for every position, and a price.
Sun to Thu 9am to 5pm · Fri 9am to 12pm · Sat closed