Their product focus is far too limited. Classification of audio is far more valuable than detecting shots. One could detect things like sentiment in a grocery store nearby a particular product (These prices are too damn high!), reasons your baby is crying, what's that weird sound with your car and who can fix it nearby, virtually any real-time, real-world analytics. It's like google analytics for audio. (Audiolytics, might I suggest). I know it has already been used to try to classify size of cars driving on the freeway.

I particularly enjoy this solution for retail locations. How long does someone talk to themselves around a pair of new pants before they buy or walk off and not buy? Is there a correlation between how long a baby screams in a grocery store and how quickly a parent leaves the store without buying something?

How about a graph showing me how long my neighbor's dog has been barking each night at 2am? How about that cough? Do other people who have coughs like that end up being diagnosed with <disease_x>?

You're assuming that their implementation is general enough to be applied well to anything other than their target market. They may use an algorithm that's only suitable for differentiating a gunshot from other banging noises, for instance.

I am. But I'm also saying that if this is the case, they should consider making some changes to generalize (or at least increase the breadth of their specialization) if they want to increase their revenue.

Their algorithm and the acoustics of the hardware, as well as the enclosure/power solution, is purpose built for triangulating gunshots sounds from rooftops in a mesh of the devices.

Integrating with public safety organizations also took a considerable amount of effort.

That is the jist of what I gathered from a former ShotSpotter employee.

I live in a neighborhood with a ton of gunshots. I've called the police about gunshots maybe three times, but I probably hear gunshots multiple times a week. One of those times I was the first person to call in a murder, the other two didn't even generate a shooting report that I could tell.

I've been following ShotSpotter, and I'm not surprised by the problem they have been having with false positives. For one thing, gunshots don't sound like cannons in real life, there are so many things that sound like them, especially in an economically deprived neighborhood. Backfiring cars, fireworks, people driving over plastic bottles, lumber being dropped, etc.

One of the reasons I don't call in shots is because it's almost impossible to figure out where a gunshot actually came from. The grid system in Chicago combined with the ubiquity of 3 story brick and stone houses means that sound bounces all over the place. Even sounds that are generally easy to place, such as trains and sirens, often sound like they are coming from a completely different direction .

The other reason I don't bother calling in shots is because I feel like any police officer close enough to respond in a timely manner is likely close enough that they probably heard the shots themselves. By the time I call 911, give them my questionable guess to location, and the police actually arrive there, chances are the shooter is long gone. Even the 911 operators seem uninterested in my calls unless I can give them extremely detailed information about what just happened.

If ShotSpotter is as accurate as the paper indicates, then I think it could be an extremely powerful tool if the information can be given to officers in a timely manner. That being said, from what I've read ShotSpotter is already considered a bit of a success in Chicago, so I think the technology has a solid market.

Given the fact that they have a solid market, I can't help but feel their problem is on the management side of things. Honestly, ShotSpotter seems like one of those companies that should have 25-50 employees, and could be reasonably profitable at that level. Looking at their SEC filing, it seems like they spend way too much on Admin and Marketing, given the fact that 1) their market seems reasonably straightforward from a sales perspective and 2) their technology seems to speak for itself.

That being said, I have no idea how much trade shows and such cost, but overall I feel like there is no reason they shouldn't be profitable after 20 years of existence given the fact that they appear to have basically the only viable product in the market.

Out of curiosity, why do you live in a neighborhood with a ton of gunshots? Can you afford to live elsewhere?

I could, but I'm saving a ton of money this way. The owner is a friend of a friend of a friend who didn't want to rent anymore, but didn't want to sell at the bottom of the market.

I live two blocks from the train, and it takes me less than 10 minutes to get to the loop. The neighborhood is also close to the neighborhoods where I like to spend time.

That isn't to say I wouldn't prefer to live in a nicer neighborhood, but I don't think the benefits are worth the added cost. Plus, to be quite frank, nobody is shooting at me.

I mean... no shit? You have a fairly complex install for a service that I don't think most police stations even want.

They have a hard enough time responding to the existing real-life humans who call them. I've reported gunshots in my neighborhood and waited more than 30 minutes for a response (we also have shotspotter).

Honestly, I wrote a wall of text about this, but I think the benefit to ShotSpotter is that it isn't a human calling it in. The lag time between the shots happening and an officer receiving information about it is so long when a human calls them in that I can see why they wouldn't even bother unless the call was extremely detailed. It isn't like people hang out after shooting at someone, and if the response takes five minutes the shooter could easy be half a mile away even on foot.

If ShotSpotter can give a location to within a couple hundred feet within 30 seconds of shots being fired, then the officers already have a much better chance of doing something useful.

I think the real opportunity would be combining data from ShotSpotter with other data sources and using them for predictive/preventative policing. That's much harder to quantify from a sales perspective though.

I think you are under-estimating the police response time by a fairly large margin. It's not like cops have shotspotter (TM) information going to straight into their heads and can react at a moment's notice. The information goes to a command center, then if a unit is available, they might send them out for a look, which means going through the dispatcher. There really isn't some police reaction force just sitting around waiting for calls.

It's dumb technology--it's like everything the MIC has done to try and re-market to police agencies, it's unnecessary and pretty much worthless.

One of the articles I read implied that ShotSpotter could send information to the computer inside the squad car. I'm not sure if that is true or not.

I'm well aware police officers aren't specifically sitting around waiting to be dispatched, but my neighborhood has a lot of idle officers that exist as a 'show of force'. I don't mean that in a derogatory way, I just mean that this officers are already dispatched to my neighborhood as a way of reducing violence. I think that responding to areas ShotTracker has flagged would be both within their capabilities in addition to being in line with their duties and priorities.

I talked to an officer in my city and they don't actually look at ShotSpotter data in real-time. They don't have the resources.

They use it to aggregate reports and consider increasing patrols (I say consider, because they often don't look at it anyway ).

Interesting. Now I'm really interested in this subject I should talk to some of the people I know in Chicago who work with policing data about it. They might have some insight.

Though I guess this is probably good motivation to go to one of the Chi Hack Nights, because I'm sure if someone there doesn't know how Chicago is using the ShotSpotter data, they probably know how to find out.

I wish they would at least open up the APIs to the public... mostly so I can use it to track down people setting off fireworks and personally go yell at them. So maybe that's a bad idea.

Some research using ShotSpotter data: https://www.brookings.edu/wp-content/uploads/2016/07/Carr_Do... . The takeaway is that gun shots are significantly underreported. Here are maps that accompany the paper: http://jenniferdoleac.com/maps

It depends on your definition of "underreported"

When you consider how often reporting gunshots or (pretty much anything) to your local PD does any good to any stakeholder I would say they're wildly over-reported.

I wouldn't doubt that guns shots are underreported. I lived in the "hood" in Milwaukee. I noticed them at first but they were so frequent that they soon blend in with the rest of the city noise. I had a new room mate move in. His first night he work me up in a panic: "What was that?!!! It sounded like gun shots!". Me: "I'm sure it was" - and rolled over. There were shots fired nearly every night. I can only guess, but I'd have to say most of those guns were not legal. I didn't once call them in. I doubt many other people did either.

Contrast that with my time living in a quiet suburb in another state. There was some public land nearby that would attract target shooters. Some neighbors were up in arms (no pun...) over it. While speaking with one neighbor she was worried about the "automatic" gunfire she could hear nearby and she was sure they were not out far enough from the neighborhood. She was right about one thing - it was in the neighborhood, but it was workers nailing sheeting on walls for a new house. They had multiple nail guns running as fast as they could. Yes - she had reported gun shots, many times. I'm sure some of them really were, but they were all very likely to be legal.

So this might just be me, but it doesn't seem creepy at all. Detecting a particular audio signal seems much less creepy than running facial analysis on everyone who walks by (https://www.facefirst.com/industry/retail-face-recognition/)

If they were recording all audio everywhere, hey, that's creepy. But it doesn't seem like that's what they're doing.

The simple fact is, you can never trust they're not doing it, no matter what it seems like or what they say. This kind of technology needs to be kept out of public places, open hardware, open source and self-hosted, without exception, otherwise it is creepy.

Actually, you can trust them.

I'm not saying whether or not that's a good decision, but saying "you can't" is misleading. Trust is a choice, and time tells whether you made the right one.

There's value in automated reporting but cities like Oakland, California, had trouble justifying the expense at current signal/noise ratios. Efforts to boost the signal are expensive and involve human analysts today. And the sales cycle with city governments can be long and labyrinthine, driving up sales costs and soaking up cash.

The business strategy is to keep revenue coming in. They need to fund R&D (and an IPO could help with that) and time to drive up signal/noise and drive down costs.

> The solution to the gunshots problem is before the guns shoot, like background checks required to own a weapon.

Most uses of a firearm in cities are by people who are already prohibited from having/buying one.

They will use stolen or straw purchased guns that are passed from prohibited person to prohibited person for as little as $50 until someone is caught with it.

Now the tech behind this actually helps the police find unreported gunshots and run reports on where most gunshots occur so they can focus their attention on that area. I don't find that troublesome in the least. As a country with high firearm ownership there are always going to be certain problems, so we try to mitigate those issues with tech.