Evaluating vendors

Ten questions to ask every firearm detection vendor, including us

This category has a documented record of overclaiming. These are the questions that settle it, put to any vendor. Every answer below ends somewhere you can check it.

Question 01

What accuracy figure do you publish, and what method sits behind it?

A federal regulator settled with a vendor in this category in 2024 over advertised accuracy claims. A number with no method behind it cannot be audited later, which is exactly when somebody asks for it.

Our answer

OptiArms publishes no accuracy figure. Each site sees its own false-alert rate instead, counted from the dismissals its own people recorded. Automated response stays locked until that measured number is low enough. An alert reports a confidence band, never a percentage.

Question 02

Where does your marketing say a concealed weapon cannot be detected?

A camera sees what is visible and nothing else. A vendor whose limits live in the contract has told the buyer one thing and the school board another.

Our answer

It is under every page on this site, including this one, and under every sales slide and press release. A build check reads every route and fails the site when a page mentions detection without it.

Question 03

After a false alert, can you reconstruct exactly why the system fired?

This is the question that arrives after the response, from a parent, a lawyer, or a board. A score nobody outside the vendor can recompute is an answer nobody can check.

Our answer

Every band is reconstructible by hand. Each sighting stores the frame, what the model scored it, what that added, and the running total. The constants that combine them are published, and a build check re-derives a whole incident and asserts the code agrees.

Question 04

Will you accept and bill a camera that cannot resolve a weapon?

Optics decide what a camera can see, and no model changes that. A vendor paid by the camera has a reason not to look too closely at the lens.

Our answer

No. Setup rates every camera against resolution floors measured from the detector, separately for rifles and handguns. A camera that resolves neither is refused, never billed, and no camera bills until video has arrived from it.

Question 05

What do you claim detection prevents, and where is the study?

The published evidence for the stronger claim is thin, and a vendor making it should hand over the paper. Detection buys seconds and warns the right people. That is worth a great deal, and it is not the same sentence.

Our answer

OptiArms does not claim to prevent a shooting, and no page here says otherwise. What it claims is detection: a visible firearm made known to the people who can act, in seconds. Every decision after that stays in human hands.

Question 06

Would your product meet a specification you did not help write?

Some state requirements in this category were drafted around one vendor's patents. A buyer can ask who wrote the wording, and whether a second vendor could satisfy it.

Our answer

Nothing here needs a specification written around it. The constants, the measured floors, and the rate card are published. A district can write its own requirement and hold this product to it alongside anybody else.

Question 07

Is your rate card public, and what notice does a price change carry?

Almost nobody in this category publishes rates, and disclosed contracts vary widely between buyers of similar size. A price that appears only inside a quote is a price that depends on you.

Our answer

The rate card is public, per camera per month, with monitoring as its own line rather than bundled. A change carries thirty days of notice under the billing terms. The amounts it replaced stay published beside the new ones.

Question 08

What does the system follow once the weapon leaves the frame?

A product that follows people is not the same product as one that follows weapons. It is the one a community objects to. Ask what is kept, who is matched, and for how long.

Our answer

Nothing. Movement is followed across frames and cameras inside a single incident, on time and place and weapon class alone. It ends when the incident does. No face is enrolled, no watchlist exists, and nobody is matched between visits.

Question 09

Can a timer or an algorithm in your product send police on its own?

An automated dispatch turns a false alert into an armed response, and the buyer inherits that decision. Ask what fires with nobody watching.

Our answer

No clock here sends anybody. A dispatch follows a human verdict, and silence is never treated as evidence. An alert nobody answers widens to the whole roster, then asks a monitoring agent to look at the frame. That ladder stops there by design.

Question 10

What happens to an alert when the reviewer, the vendor, or the network fails?

Ask this one last, and listen for the direction it fails in. A system that goes quiet when a part of it breaks has chosen the failure nobody notices.

Our answer

Review fails open. An alert nobody actions escalates on its own deadline. A watchdog reads the incidents themselves rather than trusting the queue that serves them. A cloud outage is reported to the site as a gap rather than a quiet night.

Take the list with you

Put these questions in the request for proposal as written. A vendor who cannot answer one has answered it.