Choosing CCTV analytics or ANPR software: a buyer's checklist

By Gankya · 4 October 2026 · 6 min read

Software that reads number plates (often called ANPR, automatic number plate recognition) or analyses CCTV video is sold with impressive demos. A demo, though, is filmed on the vendor's best footage. What matters is how the system behaves on your cameras, at your gates, at 2 a.m. in the rain.

This checklist is vendor-neutral. Use it to compare any options, including ours.

Start with your problem, not the features

Before talking to vendors, write down three to five real incidents from the past year and how long each took to resolve. For example: "Two-wheeler stolen from Tower 3 parking; four guards spent most of a day rewinding six cameras." These become your test cases and your measure of value. If a product cannot clearly shorten those, its other features matter less.

Questions to ask every vendor

Fit with what you have

• Will it work with our existing cameras and recorders, or do we need new ones? Will you confirm this with a site visit rather than a spec sheet?
• Which of our cameras are positioned well enough for plates, and which would need moving?
• What computer or server does it need, and where does it sit?

Performance

• How does it perform at night, under infrared, in rain and with headlight glare?
• Can it handle two-wheelers, commercial yellow plates, dirty or bent plates, non-standard fonts and plates partly hidden by a tow bar or a bumper guard?
• Can we search with only part of a number?
• When it gets a plate wrong, what does the operator see? Does every result show the image behind it so a person can check?

Accountability

• Does every operator get a personal login, with roles that limit what they can do?
• Must users give a reason before searching for a vehicle or person?
• Is there an audit log of who searched for what and when? Can anyone, including admins, edit or delete it?
• Who can export footage or data, and is every export recorded?

Data and privacy

• Where is the data stored: on our site, in India, or abroad? Who at your company can access it, and how?
• Can we set how long data is kept, and is it deleted automatically after that?
• What happens to our data when the contract ends?
• Will you sign a data processing agreement? Under the Digital Personal Data Protection Act, 2023, your site uses a processor only under a valid contract and remains responsible for it.

Support and cost

• Who installs it, who trains our guards, and in which languages?
• What happens when it breaks at night? What response time is in the contract?
• What is the full cost over three years, not just the first-year price?

Test it on your own footage

This is the single most important step. Ask the vendor to run their system on video from your cameras, or better, to connect to one gate for a trial period.

• Pick a fair sample. Include day and night, a busy morning hour, a quiet late-night hour, rain if possible, and both cars and two-wheelers.
• Make your own answer sheet. For a sample of vehicles, write down the correct number by looking at the video yourself. Do this before you see the system's results.
• Count three things: how many it read correctly, how many it read wrongly, and how many it missed entirely. Wrong reads can be worse than misses, because they send people in the wrong direction.
• Test night and infrared separately. Many systems that look good in daylight struggle after dark. Report day and night results separately.
• Re-run your real incidents. If you still have footage from a past incident, see how long it takes to answer with the new system.

Be careful with a single accuracy percentage quoted without the conditions behind it. Ask: measured on what footage, from which cameras, at what time of day, with how many vehicles?

Accountability is not optional

A tool that can find any vehicle in seconds can also be misused to follow a neighbour or an ex-partner. Good systems make misuse hard and visible:

• Personal logins, never shared accounts.
• Roles: a guard should not have the same powers as an investigator or admin.
• A reason required for every search.
• An audit log that cannot be edited, which management actually reviews.
• Exports that are logged and carry a way to check they have not been altered.

These features also protect your guards and committee if anyone later questions how the system was used.

Watch out for

• Compliance badges. We are not aware of any official certification scheme for software under the DPDP Act. Compliance is mainly about how your site uses data. Ask what a "compliant" claim actually means.
• Features you did not ask for. If features that identify people from their faces are bundled in, ask whether you can switch them off, and whether you need them at all.
• Lock-in. Can you export your data in a standard format if you leave?
• Hidden cloud. "On-premise" sometimes still sends data elsewhere. Ask exactly what leaves the site.

Run a proper pilot

• Agree success measures in writing before starting, for example time to trace a vehicle, read rates by day and night, and guard feedback.
• Keep it focused: one gate or one phase, for a fixed number of weeks.
• Make sure your guards use it during the pilot, not just the vendor's engineer.
• Agree what happens to the data at the end of the pilot.
• Ask for a written results report, including what did not work.

Work out the total cost

Add up, over three years:

• Software licence or subscription, plus GST.
• Hardware: computer or server, storage, any extra or replacement cameras.
• Installation, camera repositioning, cabling and power backup.
• Training, including for new guards when staff change.
• Annual maintenance and support.
• Your own staff time to run and review it.

Compare that against the hours your team spends on incidents today, and against the cost of incidents that are never solved.

Gankya is the intelligence layer on top of the CCTV system you already have: it connects to your existing Hikvision, CP Plus, Dahua or ONVIF recorder, and the recorder keeps the video. It is pre-pilot: it works on our demos and test video, the recorder connection is tested against simulated recorders, and we have not yet measured it on a real site or at night under infrared. We verify your recorder in the site check and pilot. That is exactly why we start with a 30-day pilot on your own cameras, with a written results report. If you would like to put us through this checklist, request a pilot.