A practical, formula-first guide: the bitrate-driven math behind NVR storage calculation, the variables that move it, and how to set retention so you buy the right capacity once.
Start with the formula
The fastest honest answer to any NVR storage calculation is the same equation every integrator uses:
Storage (GB) = (Bitrate in Mbps ÷ 8) × 3,600 × hours/day × cameras × retention days ÷ 1,000
Dividing the bitrate by 8 converts megabits to megabytes; multiplying by 3,600 gives you an hour; the rest scales the hour out across your camera count and how many days you keep footage. As a rule of thumb, one camera streaming a steady 4 Mbps for 24 hours produces roughly 43 GB per day, or about 1.3 TB over 30 days. Multiply that by your channel count and you have a defensible first estimate before any vendor quote enters the picture.
The trap is treating bitrate as a fixed number. It is not. A 4K camera staring at a busy loading dock behaves nothing like a 2 MP camera watching a quiet stairwell, and the spread between them can be 5x or more. Sizing well is mostly about pinning down bitrate per scene, then layering retention and headroom on top.
Lock down the five variables that move bitrate
Before you calculate, get honest numbers for each camera or, more practically, for each camera type in your design:
- Resolution. 2 MP (1080p), 4 MP, 5 MP, and 8 MP (4K) each roughly step up the data load. More pixels, more bits.
- Frame rate. 15 fps is fine for general overview; 30 fps doubles many scenes and is worth it only where you need to capture fast motion (POS lanes, gaming floors, license-plate capture).
- Codec. H.265 typically cuts bitrate 30–50% versus H.264 for comparable quality, and vendor-specific smart codecs (zipstream-style variable bitrate) cut it further by spending bits only where there is motion.
- Scene complexity and motion. Foliage, crowds, traffic, and rain all raise bitrate because the encoder has more change to describe frame to frame. A static wall is cheap; a windy parking lot is expensive.
- Lighting and noise. Low-light scenes generate sensor noise that the encoder faithfully records, inflating bitrate at night precisely when you assumed it would drop.
Pull real bitrate figures from the manufacturer's storage calculator for your exact models, then sanity-check against a short live test on a representative camera rather than trusting datasheet best-case numbers.
Choose recording mode honestly
Recording mode is the single biggest lever after bitrate, and it is the one most often fudged.
Continuous recording captures every second. It is the only mode that survives serious scrutiny in incident reconstruction and is the safe default for anything evidentiary or compliance-driven. Size it at the full 24 hours per day.
Motion-based or event-based recording only writes when something triggers. Vendors love to quote 50–70% storage savings here, and in a genuinely quiet scene that can be real. But it is a planning hazard: a swaying tree, headlights, or a poorly tuned detection zone can push effective recording time far past your assumption, and a missed trigger means missing footage exactly when you need it. If you must model motion savings, be conservative, estimate 12–16 active hours rather than 4, and never apply aggressive motion assumptions to perimeter, cash-handling, or egress cameras.
A common professional pattern is hybrid: continuous at a modest frame rate as a baseline, with higher frame rate or higher quality triggered by events. Size the continuous baseline as your floor so you are never caught short.
Set retention to the requirement, not the guess
Retention days drive cost linearly, so anchor them to an actual obligation instead of a round number:
- Find the governing requirement. Many commercial sites land at 30 days. Regulated environments often run longer, and some federal, financial, or critical-infrastructure programs specify 60, 90, or more. Check the contract, the regulation, and your insurer before you pick a number.
- Add an investigation buffer. Incidents are frequently discovered days after they happen. If your policy says 30 days, the practical floor is often 30 days of guaranteed retention, not 30 days that quietly shrinks as bitrate spikes.
- Consider tiered retention. Keep all cameras at full quality for the mandated window, then optionally down-sample or retain only flagged events longer. This controls cost without violating the core requirement.
Write the chosen retention into the system design document so it survives staff turnover and the next refresh cycle.
Add headroom, RAID, and real-world overhead
Raw footage size is not usable capacity. Three adjustments separate a clean spec sheet from a system that actually holds your footage:
- Fill ceiling. Never plan to fill a recorder to 100%. Size to about 80–85% so the array has room to breathe and performance stays stable.
- RAID overhead. Redundancy costs capacity. RAID 5 sacrifices one drive's worth; RAID 6 sacrifices two but tolerates a second failure during rebuild, which matters with today's large drives and long rebuild times. Factor the parity loss into usable capacity, not nameplate.
- OS, database, and growth. Reserve space for the VMS database, logs, and the cameras you will inevitably add. A 10–20% growth allowance now is cheaper than a forklift upgrade in eighteen months.
For larger deployments, separate the recording and management roles: the VMS coordinates, searches, and exports while one or more recorders hold the bulk video. This is also where you decide between an appliance and server-plus-storage, and where network design (multicast, bandwidth to the recorder, failover) starts to matter as much as the drives themselves.
Validate, then verify the supply chain
Run your NVR storage calculation at least twice: once with conservative datasheet bitrates and once with the manufacturer's calculator using your actual resolution, frame rate, and codec. If the two diverge sharply, find out why before you buy. Then pilot a few representative cameras for several days and measure real consumption against your model. The delta you find is the number that would have hurt you at scale.
One step that is easy to skip and expensive to ignore: the recorders, drives, and cameras all carry supply-chain obligations. For federal and many enterprise buyers, video surveillance equipment must clear NDAA Section 889 prohibited-source rules, and GSA-track purchases must meet TAA country-of-origin requirements. Compliant VMS and camera platforms exist across the market, but a single rebranded NVR or surveillance-grade drive can put an otherwise clean project out of bounds. Confirm the country of origin and 889 status of every SKU, not just the camera brand, before the purchase order goes out.
The bottom line
Good storage sizing is unglamorous arithmetic done with honest inputs: real bitrates, a recording mode you can defend, a retention window tied to an actual requirement, and headroom for redundancy and growth. Get those four right and the capacity number takes care of itself, vendor-neutral and refresh-ready.
If you want a sizing model built around your camera counts, scenes, and retention obligations, and validated against 889 and TAA from day one, request a quote scoped to your storage and retention requirements.