A city’s request for proposals (RFP) tells you what gets installed. It rarely tells you what will still be working in year five, after the budget, the staff, and the compliance rules have all changed.
A City Is Not the Buyer the Spec Assumes
For most integrators bidding public-sector work, the RFP is gospel. Camera count, resolution, lens type, storage in days, sometimes a named video management system (VMS) with “or equal” attached. The response comes back as a compliance matrix, yes meets, line by line, because nobody on the bid team talked to the people who will actually use the system, only to whoever wrote the RFP. It answers the spec instead of the requirement.
Across northern California and Nevada, for example, the same gap is visible from the manufacturer’s side of those projects. The spec tells a city what gets installed. It almost never tells them what keeps working.
The other thing a spec rarely accounts for is who it is describing. A city is not one organization. Most commercial accounts have one security team, one policy, one org chart. A city is federated; police, transit, parks and recreation, utilities, courts, each with its own legal authority over its own footage, its own retention obligations, and its own rules about who is allowed to see what.
Then there is what happens after the install. Who maintains the system. Whether it scales when the city adds fifty cameras in year three without opening a new procurement cycle. Whether the platform survives a firmware change without replacing edge devices. Year-five failure is almost never “the cameras stopped working.” It is that nobody left in the building knows how the system is licensed, integrated, or backed up. That is a governance question, and the spec sheet never asks it.
The Requirements a City Does Not Get to Choose
The spec is not even where many of these projects start. More municipal conversations now begin with a compliance mandate than with a security problem. NDAA Section 889 comes up early for any agency with a federal funding nexus, and it forces a rip-and-replace conversation on banned manufacturers whether or not the buyer had budgeted for a refresh. In California, for example, state procurement rules layer on top of the federal mandate, so the compliance discussion can begin before anyone asks the security team what problem they are trying to solve.
A conversation that starts with compliance can just as easily end with it. The risk is treating the mandate as the finish line rather than seeing the whole system. Swap a banned camera for a compliant one, check the box, stop there, without revisiting whether the VMS, the retention plan, or the integration story still makes sense for the next decade. Compliance is the floor, not the whole spec. If a rip-and-replace is going to happen anyway, that is the moment to fix the architecture, not just the hardware.
Federal mandates are the most visible of these rules. Three more arise from outside the security team, and none of them are open to negotiation:
- Each department holds legal authority over its own footage, so the platform has to keep them apart and prove it did. Role-based access and segmentation granular enough that a transit supervisor can never pull up a police interview room. An audit trail solid enough to prove that segmentation held when someone challenges it in a records dispute or in court. In California and Nevada, someone eventually does.
- Retention is set by state public records law, department policy, or statute tied to the footage type: body-worn, in-custody, traffic stop. It varies state to state and department to department inside the same city. Agencies often get this wrong, usually because whoever wrote the RFP pulled a number from a template instead of checking their own state code. The number must come from records or legal before storage is architected, with headroom built in, because these laws get amended more often than people expect, almost always toward longer retention.
- Public records requests are messier than they look in a sales conversation. Footage has to be located, then reviewed and redacted: faces, plates, bystanders, anything outside what is responsive or exempt under the California Public Records Act or its Nevada equivalent. Locating it is harder than it sounds when departments are not unified on one platform. Automated redaction helps, but a person still has to review and certify the output before release. This rarely gets scoped before purchase; most agencies find out how bad it is the first time a high-profile request lands on someone’s desk.
Public Budgets Change What a Licensing Model Has to Survive
Public budgets do not work like corporate ones, and licensing models built for the private sector often do not survive contact with them. A subscription or annual license is an operating expense that has to clear a budget vote every year. A capital purchase is a one-time appropriation, politically easier to get through a council. Grant funding carries a spend-by deadline and restrictions on use, which pushes buyers toward capital-heavy, perpetual-license structures even when a subscription would serve them better technically. In the Bay Area alone, Urban Area Security Initiative and California Governor’s Office of Emergency Services money each runs on its own clock.
The workable approach is to design the full architecture up front but phase the purchase to match fiscal years and funding sources, so a grant-funded phase one does not box the city into hardware or licensing that a differently funded phase two cannot build on. An open platform earns its keep here for a practical reason: it lets phase two be bought with different money, on a different clock, from a different vendor, and still land on the same system.
What Stalls a Municipal Project Is Rarely the Technology
Phasing a purchase assumes you know how long the project will take, and municipal timelines almost always run longer than the technology. Consider a mid-size deployment that takes three years to close. The integration work is usually the small part of it. Access control may need only a shared credential database rather than a deeper API integration, enough to correlate badge events with camera footage. Dispatch integration is often deliberately minimal: location data so footage can be tied to an incident address, with no bookmarks or alerts pushed into computer-aided dispatch (CAD). On existing recorders, the sensible direction is to keep whatever can be brought into the new platform and swap out only what cannot be integrated or made compliant.
The delay is organizational. A procurement touching police, IT, facilities and often more needs every one of those departments to weigh in on requirements, budget and prioritize before anything moves, and aligned on a single purchase is where the timeline actually lives. Board approval schedules and fiscal-year budget cycles set the pace, not the install calendar. An integrator bidding municipal work at commercial-sector speed loses credibility with the buyer the first time the calendar slips, and it will not be the technology’s fault.
Cloud Rules Are Moving Faster Than Agencies Are.
The technology does change , and cloud is where it is going now, slower than the marketing suggests but with real momentum behind it. StateRAMP officially became GovRAMP last year, though everyone in the industry still calls it StateRAMP out of habit. It is real, and it is getting teeth fast: Nevada now requires GovRAMP authorization for executive branch cloud contracts as of July 1, 2026.
Even so, plenty of agencies still do not have a GovRAMP-authorized option that fits their use case, or their internal policy has not caught up to permit cloud storage of evidentiary video at all, mandate or not. Institutional caution is doing real work here too. Where cloud is technically permitted, public safety and IT leadership remain conservative about the chain-of-custody implications of footage living outside a facility they physically control.
Nevada is worth watching, though. A hard mandate in a single state tends to force that conversation faster than the market would on its own. Until it does, hybrid is what bridges the gap, and it is the honest current state: cloud for management and alerting, on-premises or hybrid storage for anything that might end up in court. That split follows the caution rather than fighting it.
Designing for the Day the Champion Leaves
Whatever the deployment model, it will outlast the people who chose it. Administrations change, department heads rotate, and the person who drove the project moves on to something else. A deployment survives when the business case lives in the documentation rather than with a person: the rationale, the standard operating procedures, and the training written into policy, so that a successor inherits something that explains itself.
The most durable thing an agency can do is tie the platform decision to its own mission. A system remembered as one leader’s preference is easy for a new administration to set aside. One documented as meeting the agency’s retention, interoperability, and evidentiary requirements carries its own justification, because those requirements outlast the people who wrote them.
Where Municipal Video Is Heading
What agencies are asking for is changing too. Real-time crime centers are the clearest growth trend. Mid-size cities are standing up centralized monitoring that fuses camera feeds, automated license plate recognition, and dispatch data, a capability that used to be limited to large metros. Camera-sharing programs with local businesses are expanding too, mostly because they extend coverage without capital spend.
The trend worth watching most closely is community oversight. Several cities have passed surveillance oversight ordinances requiring public disclosure of what is deployed and audit rights over who accessed footage and when. That pushes audit logging and access reporting from a nice-to-have into a baseline requirement, and it changes what has to be demonstrated at the time of sale rather than at the first audit.
What Separates the Projects That Work
Everything here points the same direction: a municipal video system is bought once and lived with for a decade, and the spec only describes the first day of it. Who is permitted to see what, what the statute requires, where the next phase’s money comes from, who can explain the design after the people who chose it have moved on: none of that appears on a compliance matrix, and all of it determines whether the system is still doing its job in year five.
That makes this less a procurement problem than an institutional one, which is also why government work rewards a longer view than commercial work does.The agencies that get the most from their systems, and the integrators who stay with them longest, are the ones who answered the requirement rather than the spec, who treated the purchase as a decision about how the city would operate, not about what would be installed.
