Feb. 21, 2020
The LiveSafe Platform is about to get smarter — a lot smarter.
This year, we will be releasing LiveSafe Insights, the industry’s first artificial intelligence and natural language processing capability to classify and categorize crowdsourced tip data, intelligently guide users through the tip submission process, and help organizations better understand their risk profile.
By applying advanced algorithms to seven years of tip data, LiveSafe is now capable of showing organizations how their LiveSafe deployment and risk management program is performing compared to their peer organizations. Natural language processing has also demonstrated the ability to improve the categorization of tip data and unearth unknown risks to an organization.
Last week, I interviewed Dan Morrison, LiveSafe’s Senior Director of New Products, about LiveSafe Insights and the new capabilities that it will bring to LiveSafe clients and their risk management programs.
Q. Dan, this is going to be a pretty big year for LiveSafe. Can you tell us what LiveSafe Insights is and where the concept for developing Insights came from?
A. I agree. Very exciting year. LiveSafe Insights is an analytics engine that we are layering on top of our LiveSafe offering and basically, what we’re trying to do is parse through our client’s data to give them information about their unique risk profile that they face on a day-to-day basis and how they can best address that, as well as helping them maximize their use of the LiveSafe product using a number of recommendations and other sorts of interventions that we see that will help them best utilize the product.
Really what this came from is I oftentimes divide products in the space into two different camps. There is generally the react and respond, which is oftentimes about triaging and event management. Really the DNA of LiveSafe is primarily about detecting and preventing. What I love about the product is that it actually circles through our client’s data and ultimately, it’s able to use a lot of interesting algorithms that we’ve been developing to detect certain types of incidents and ultimately, surface whether they’re trending up, whether they’re trending down, whether they’re related to other aspects of your community and LiveSafe deployment and ultimately, allows people to better align their resources to address those.
Q. Can you describe for our audience the technology that powers this analytics engine?
A. Probably the most valuable part of this product is in fact the LiveSafe Platform as it exists today. We’ve been deployed all across multiple different industries from healthcare, to arenas to utilities, to higher education and so we’ve got a pretty rich data set that we’ve actually been able to work from, and from that dataset we’ve actually been able to build both supervised and unsupervised models in order to classify, cluster, and otherwise group some of these incidents that we’re detecting. What that enables our clients to do is to benefit from the global deployments that we have with LiveSafe. If you’re a higher education client, you very well can benefit from the sorts of technologies and algorithms that we’ve developed for our arena clients and our healthcare clients. Ultimately, what we’re looking at is using a lot of natural language processing of the information that comes in from the hearts and minds of your employees. Being able to group that, being able to classify it, being able to triage it into the right places of the organization, giving you a lot more situational awareness about what you’re seeing in LiveSafe and what you can do about it.
Q. What would you say were some of the specific challenges that companies and universities were facing that this new LiveSafe Insights product will help solve or address?
A. I think a big portion of what we have to offer is surfacing information that in my mind comes from the world’s most valuable dataset, which is the people that you work with in your professional community or your educational community. We oftentimes see portfolios of tips that come in and necessarily being able to turn those into valuable intelligence or actionable intelligence is not a straight forward solution. There’s a lot of different ways that you can score this data. There’s a lot of different ways that you can bucket this data. There’s a lot of different ways that you can tag and classify this data.
Insights has been our foray into taking our best learnings across a wealth of different industries and applying that sort of algorithmic rigor to all of our client’s data sets. What that ends up allowing us to do is having better deployments of LiveSafe that generate more data, that generate more relevant intelligence and help you know exactly what to do with it. Gone are the days where you would have to go through, read every single individual report and figure out where does this fit. We’ve developed a number of frameworks to organize and taxonomize this information and then give you a digestible readout of what we’re seeing and what you might want to consider because of that information, which moves us much closer to the idea of detecting, preventing, and doing more with less from a safety perspective.
Q. I also think it’s important to note that one of the main differentiators here is the fact that we’ve been at this a long time, we have seven years of data, so we’re not starting from scratch, right?
A. Exactly. I think one of the great benefits that we have had as a company and one of the great joys of working on an offering like Insights has been that we’re able to mix and match different findings that we found across a number of different industries. Every industry actually benefits from the host of different deployments that we have. I think that that makes it a very unique offering in that when you use this product, you’re not just buying into your direct vertical, you’re buying into the intelligence, the learnings, the findings, and the analysis that we’ve deployed for other clients to be successful. Ultimately, what you’re looking at is cross pollinating best practices across all of our LiveSafe ecosystem of clients. That enables the product to be a lot more resilient, detect a lot more anomalous behaviors that would not be possible if we were focusing on an individual industry and it gives us a pretty fantastic testing ground to vet different analyses that we’d like to run. When they win and when they do well, we can roll those out to our entire portfolio of clients.
Q. I think LiveSafe Insights is really going to change how organizations of all shapes and sizes look at the safety app market. Do you agree? This is really exciting, right?
A. To say that that I’m excited about it would be a huge understatement in the sense that it allows us to move beyond just ingesting information. For the most part, submitting a tip, it’s a pretty straightforward process. You write the text in, maybe you categorize it, maybe you don’t and then someone triages it and sends it to where it needs to go. The idea of beginning to move to useful portfolios to tell you unknown things about your business that live inside the hearts and minds of your employees and being able to actually guide your prevention response and develop your prevention strategy, based on not just one tip or two tips, but your tips and adjacent LiveSafe deployment tips, because we’re able to peer and benchmark against other participants in the Insights community. That is what allows us to actually offer a higher caliber service that goes from just tips coming in to dealing with a lot of algorithmic rigor that allows you to do a lot more with a lot less, and also allows you to get a much smarter use out of a platform that ultimately, can save you dollars and most importantly, save you harm and lives of your employees.
Q. Now there’s something else that our natural language processing capabilities provide and that is help to the app user when it comes to knowing how to categorize a tip and what to report. Can you talk to us a little bit about that aspect?
A. Right now, what we’re able to do is actually during the tip submission process, as a user is submitting information to their organization, we can make recommendations based on what we’re detecting in the language that they’re using to help them make sure that they are classifying that information in a way that it gets to the right place, at the right time, and they’re able to get the help that they need. That helps the user get more confidence that what they’re reporting is being understood by the system. It’s being understood by my employer or my university or this arena, whatever client deployment it might be, the user feels understood. Instead of a user having to guess which department does this go to? What are the circumstances in which I may use this category or this category? LiveSafe can intelligently guide the user along to make sure that they get the right classification onto a tip so it gets to the right place at the right time.
That ultimately helps users to feel a little more confident in the solution, that [their tip] is going to get heard, it’s going to the right place and that leads to better reports. It leads to higher volumes of reports and with better and higher volumes of reports, that actually gives us more data to refine that recommendation algorithm so it gets better over time. One of the things that we’ve realized is, we’re able to take [a client’s] uncategorized tips and unmask what’s behind that. We’re able to recategorize that portfolio and we can tell them, ‘Did you realize,’ for example, ‘that you may actually have high instances of people reporting facilities concerns in this building as opposed to what you initially purchased LiveSafe for, which may have been traffic and parking concerns or sexual assault concerns or drugs and alcohol concerns.’
We can actually make recommendations on what the users are choosing to submit information on rather than what the client is hoping that they’re going to report. In developing the product, we actually were working with a client and one thing that the client had not provided in their deployment was the ability to report sexual assaults. Their take was, ‘I didn’t purchase LiveSafe for sexual assaults and I don’t really have that problem at my organization.’ After overlaying a bit of algorithmic classification on top of their portfolio of tips, we actually detected quite a few tips that had language that suggested this is a sexual assault tip. We were actually able to go back to that client and say, ‘Not only did you potentially miss these'” because they had pretty high volumes of tips coming in, ‘but actually it suggests that this may be a channel that you should calibrate in order to pick up these sorts of sexual assault tips.’ What otherwise would have been passed over, we were able to scan on behalf of the client, go in and make recommendations and able to further tweak their LiveSafe deployment to make sure that it could cover an even broader base of risks than it initially was purchased for.