The useful answer is rarely “buy the biggest number.” This guide turns the specification sheet into an installation decision you can defend after dark, after handover and when footage is actually needed.
Measure bitrate instead of guessing
Retention calculations are only as reliable as the bitrate assumption. Resolution alone is not enough: frame rate, scene movement, noise, codec and quality targets all change the amount of data. Use the camera's measured average and peak bitrates from a representative day, then calculate each stream individually.
Model motion and difficult scenes
Motion recording can save space in quiet locations but should never be treated as a fixed percentage without testing. Rain, trees, insects and changing light can keep a scene active all night. Smart detection reduces nuisance recording only when zones and object filters are correctly tuned.
Turn retention into a policy
Choose surveillance-rated drives, allow space for database overhead and avoid operating an array completely full. Retention is a policy decision as much as a technical one: keep footage only as long as there is a legitimate need, protect exports and document who is allowed to retrieve them.
Commission the system against a written evidence goal. Save reference images and settings so future maintenance can spot gradual changes.
Can another person prove the system still works?
Record the final view, night image, bitrate, alerts, firmware and access method. A system is not finished until the owner can verify recording and export without the installer standing beside it.
Every site is different. Confirm manufacturer instructions, electrical requirements, privacy obligations and local regulations before installation.
Frequently asked questions
How is NVR storage calculated?
Add the measured average bitrate of every stream, convert bits to bytes, multiply by recording hours and retention days, then include operational headroom.
Does motion recording always save CCTV storage?
Not reliably. Rain, insects, trees and lighting changes can keep a scene active, so motion assumptions must be tested in the real location.