Report 072 · Luxury Home Security
You can spend six figures on an estate security system and, in a growing list of American cities, no officer will be dispatched when it goes off. This is not a budget failure or a slow night. It is written policy, it has been running for a quarter century, and the reason is a number almost nobody who buys an alarm has been told.
Every alarm sold in this country comes with an implied promise, and the U.S. Department of Justice put its finger on exactly what is odd about it. Purchasers "are told to expect a police response to an alarm activation, even though they bought the system from a private alarm company with no link to a police department."
That promise is being withdrawn, city by city, and it has been happening since 2000. If you own a large property, the question is not whether your system detects an intrusion. It is whether anything your system produces counts, in the specific legal sense used by your city's ordinance, as verification.
The definitive treatment is a U.S. Department of Justice guide, Rana Sampson's False Burglar Alarms, published through the COPS Office and the Center for Problem-Oriented Policing. It is not new, and I will come back to that, but its central finding has never been seriously disputed:
"The vast majority of alarm calls—between 94 and 98 percent (higher in some jurisdictions)—are false."
Turn that around, as the guide does, and you get the sentence that should be on the first page of every alarm contract: "alarms' reliability, which can be measured by these rates of false activations, is generally between 2 and 6 percent."
The scale follows from the rate. In 2002 U.S. police responded to roughly 36 million alarm activations at an estimated annual cost of $1.8 billion. False alarms accounted for somewhere between 10 and 25 percent of all calls to police nationwide, and for many agencies they were the highest-volume call type there was. The guide quotes an estimate that solving the problem "would, by itself, relieve 35,000 officers from providing an essentially private service."
Individual cities are worse than the average. Of Dallas's 62,000 alarm calls in 2004, 2.8 percent were valid. In Salt Lake City in 1999, of thousands of alarm calls answered, 0.3 percent resulted from crime. Eugene, Oregon logged 5,944 alarm calls in 2001 and made ten arrests.
Verified response is the policy that follows. Under it, alarm monitoring companies "must verify the legitimacy of alarms (except holdup, duress, and panic alarms) before calling the police," and verification "typically involves visual on-scene verification of a break-in," or remote video surveillance, or private security personnel who travel to the site and assess it.
Sandy Springs, Georgia is a useful live example, because a well-off Atlanta suburb is precisely the market this beat covers. The city's own description is unambiguous: "an intrusion alarm must be verified by audio, video, or private guard response before police dispatch will occur." Its definition of verified is "visual or audible confirmation of an attempted or actual crime, fire, or other emergency situation at the alarm site." The standard took effect on 19 June 2019 after a year of implementation time.
The city's stated reason matches the national picture: "more than 99 percent of calls from monitoring alarm companies are false alarms, and approximately 18 percent of calls received by the City's 911 Center originate from alarm companies."
Read the definition slowly, because it is doing something specific. It does not ask whether a sensor fired, whether a detection algorithm was confident, or whether the monitoring center believed something was wrong. It asks for confirmation of a crime. Those are different claims, and only one of them is a fact about the world.
The part that gets lost in every "police stop responding to alarms" headline is that they do not stop responding to everything. Sandy Springs police state it plainly: "Public safety personnel will always respond to panic, duress, hold-up, fire, and medical alarms as well as to 9-1-1 calls from a person at the alarm site."
The DOJ guide carves out the same categories, for a reason it states directly: when a human being personally pushes a button, "gaining additional verification before dispatching a police officer is unrealistic."
So the operative distinction is not detection quality. It is whether a person asserted an emergency or a machine inferred one. A perimeter beam, a glass-break sensor and a motion analytic are all machine inferences. A duress code, a panic pendant inside the house and a phone call from a resident are all human assertions, and they are all still on the fast path.
That has a concrete design consequence for a large property, and it is the opposite of what most people spend money on. Adding a fourth camera to the driveway does not change your response tier. Making sure the people who are actually on the estate at 2 a.m., including staff, can make a human assertion that reaches dispatch, does.
Every vendor in this market now sells AI verification, usually with a claim about reducing false alarms by some large percentage. I want to be careful, because that technology is real and video verification genuinely works. But the industry has run this experiment before, on a modality that sounded exactly as convincing at the time, and the results are worth putting on the record.
Audio verification was supposed to solve this. An operator listens to live audio from the site and decides whether to call the police. Several agencies measured it. London's Metropolitan Police found an 80 percent false rate. Fremont, California found 96 percent across a year of audio alarms. Salt Lake City's own police measured 82 percent over several years. Burien, Washington found 92 percent across nearly seven years of calls. The guide's verdict on audio is blunt: not nearly as effective as visual on-scene or video verification.
This is the part of the problem I know from the other direction. In counter-IED and electronic warfare you spend your professional life on the distance between a detection and a confirmation, and on what it costs to treat the first as the second. The lesson is always the same, and it is a measurement lesson before it is a security one: a detector's sensitivity, quoted alone, tells you nothing you can act on. What matters is how often the thing that fired was real, in this environment, at this alert rate. A sensor that is 99 percent accurate in a setting where genuine events are rare still produces an alert stream that is mostly noise. That is not a flaw in the sensor. It is arithmetic, and no amount of model quality repeals it.
Which is why an ordinance is a more honest instrument than a spec sheet. A confidence score is a claim the system makes about itself. An ordinance grades the output: either something confirmed a crime at the site, or it did not. Vendors are measured on sensitivity because sensitivity is what demos well. Cities measure the false-dispatch rate because that is the number they pay for.
The economics are the most underreported part of verified response, and Sandy Springs made an unusual choice worth studying.
Fines there are assessed to alarm companies, not to homeowners. The city says so directly: "False alarm fines are assessed to alarm companies, not individual alarm users." The police department adds the mechanism, that companies are fined "when they request public safety response to an alarm activation that turns out to be false and the cause of the false alarm is not user error," and then the honest caveat, that the company "may or may not pass false alarm fines on to its customers." Registration follows the same logic: alarm companies register their customers' systems rather than homeowners registering themselves.
Put the incentive where the decision is made and behavior changes. The entity choosing whether to call the police is the monitoring center, so that is the entity holding the bill. Compare that to the conventional permit-and-fine model, where the homeowner pays for a dispatch decision somebody else made using equipment somebody else installed. The DOJ guide lists the actual causes of false alarms as user error, faulty or badly chosen equipment, and poor installation. Two of those three are not the homeowner's doing, and one UK study cited in the guide put user error at about half of activations, which leaves a great deal of it sitting with the industry.
Two more findings that cut against the scare stories. Salt Lake City adopted visual verified response in 2000 and the police department gained the equivalent of five full-time officers, reduced call-taker and dispatcher workload, and improved response times to other calls. Meanwhile "area alarm industry representatives cited increased revenues (as a result of the service charge applied for verification) and similar sales levels to those before the verified response policy." And on the insurance question that every homeowner asks: in cities adopting verified response, insurers kept giving discounts, "as it is the monitoring itself, not whether it is done by police or private security, that appears to matter."
Cities adopting the policy saw alarm call volumes fall by roughly 90 percent. Not because burglary fell. Because the calls were never burglaries.
Three practical consequences follow, and none of them are about buying a better sensor.
First, find out what your jurisdiction actually requires before you specify anything. Verified response is a local ordinance, and neighboring towns differ. The question to ask your city, in writing, is which specific forms of verification trigger dispatch.
Second, if you are in a verified-response city, private guard response is not an upsell, it is your response tier. The DOJ guide is candid that this is where the model strains: jurisdictions with few alarm customers spread over a large area "may have difficulty securing a private resource that can deliver satisfactory and cost-effective response times." An estate on acreage outside a service density is exactly that case, and the honest answer is that a contracted responder with a real time commitment is the thing being purchased, not the panel.
Third, design for the human-assertion path. That is the one the ordinance still honors without argument.
I should say where I sit. I help design the AI security systems for a veteran-owned (SDVOSB) luxury home-security company run by fellow veterans. I do not own it and earn nothing from this link; I flag it because it is a field I build in and not only write about. It is also why I am wary of the marketing: the honest version of what good analytics buy you in a verified-response city is not fewer false alarms in the abstract, it is a recorded image a monitoring operator can look at and say a person is on the property right now. That is what the ordinance is asking for. Full policy here.
The DOJ guide is the second edition, published in August 2011, and its headline national figures date from the 2000s. I am citing it because it remains the most rigorous public treatment of the problem and because its structural findings have held, but I am not going to pretend a 2002 activation count is a 2026 activation count. If someone has published a comparable national measurement since, I did not find it.
The same caution applies to adoption. The guide reported that between 20 and 25 U.S. cities had adopted verified response as of its writing, along with several Canadian agencies. I did not verify a current list, and you should not assume your city is or is not on it. Check locally.
The audio-verification false rates above are all from 2006 agency measurements. They are a legitimate warning about how confidently a verification modality can be sold before anyone measures it in the field. They are not a measurement of any camera or algorithm sold today, and I am not presenting them as one. The guide itself rates visual and video verification substantially higher than audio.
Finally, this is contested policy, not settled consensus. The guide records that most agencies adopting verified response "had to withstand significant resistance from the alarm industry," that the industry has defeated the proposal in many other cities, and that the International Association of Chiefs of Police position, developed with alarm industry bodies, favors telephone or electronic verification rather than the mandatory visual standard. The common objections are that the public expects a police response and that police are better trained than private security for these calls. Both are reasonable, and neither makes the 94 to 98 percent go away.
An alarm system is a detector, and for a hundred years we have been buying detectors and calling them protection. The distance between those two words is the entire subject.
What verified-response cities did was refuse to keep absorbing that distance at public expense. They did not say alarms do not work. They said an alarm activation is a hypothesis, somebody has to test it before an armed response is dispatched, and it is not going to be us.
If you own a property worth protecting, that is not bad news, but it does change what you are shopping for. The valuable part of your system is no longer the sensor that notices. It is whatever can stand behind the sensor and say, to somebody with authority to act, that this one is real.
Scope note: verified response is set by local ordinance and varies by jurisdiction. Nothing here describes the policy in any city other than the one named, and it should not be relied on as a statement of what your own city requires. The DOJ guide's national statistics are from the 2000s and are reported here as of their publication date rather than as current figures; the same applies to its count of adopting cities and to all quoted audio-verification false rates, which are 2006 agency measurements and are not measurements of any product on sale today. The comparison to detection and confirmation in counter-IED and electronic warfare is the author's own professional background, offered as an analogy, and is not sourced to any cited document.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.