 I know if one thing was reinforced, listening to all our speakers today, it's that actionable insights are vital. And that's something we really think about at Storyful. If our insights end with a presentation, or they end with a meeting and there are no action points after that, we've failed. That's not to say presentations aren't great, but there needs to be a next step. And if you're not providing actionable insights, something's gone wrong. So I'm going to start off just by talking a little bit about crisis management. And a crisis manager might not be a role you're too familiar with. So I tried to break it down really simply. This is not a crisis. This is the monitoring stage. And crisis managers spend most of their life in this stage where it's almost like institutional paranoia, always watching, always wondering if they're missing something. And then at some point it becomes a crisis. Now, for us at Storyful, our job is to ease that pain. And we do this through a number of ways and we use a lot of data sources to do it. But I want to talk about a very specific type of crisis because crisis can mean a lot of things to a lot of different people. For us we're specifically talking about reputational crisis here. That is a crisis where your consumer base or even the wider audience has lost faith in your brand. Now, with the advent of social media, this has become a massive issue for big companies and small companies. And if you think of reputational crises, it breaks into kind of two stages. And it's almost, it's a little bit like a pathogen. The first kind of the starting point is often an issue, which is a non-acute risk to an organization's interests. And these can fly under the radar or they can be very well known issues that a company is not in the place to resolve or frankly they just aren't interested in it. If you think of incidents then, that's almost like the symptoms. This is the outbreak. And this is where a company suddenly realizes we've messed up here, something has happened. It's by nature unplanned. If it was planned, we would be in serious trouble. And it can either evolve from an issue that's already known or it can happen completely separately. Now, at this incident stage, in order to mitigate against it, you do require a quick or targeted response or else you may face financial or permanent reputational loss. So I won't read out the definition of a crisis management. Basically, that's working quickly to mitigate against issues and incidents. So roughly speaking, issues and incidents can be divided into internal issues and incidents and external issues and incidents. It's actually inaccurate to put these in a quadrant because they shift, they change, they overlap, but it was the easiest way just to present this. So giving you some examples of internal issues, these are often the ones that lead to reputational incidents. You've got your poor culture, poor security practices, policy flaws, malpractice habits, spotty communications, a big one, or insufficient training. External issues then are the bigger problems that face many corporations at once and that can include infrastructure problems in the country, regulation, environmental problem, politics shifts, economic changes, and cultural shifts as well. So then, as I said, all of these can change and kind of swap and I wish there was a nice way to kind of predict what was going to happen. But these are some of the classic internal incidents. You've got your lawsuits, your accidents, HR issues, product defects, which can be also an external incident as well. Same with hacking and one of our favorite is disgruntled employees. Then, if you look at your external incidents, these can be stuff like GDPR and Uncle Blinds, pollution, exposés, boycotts, or wider perception problems. So this is your classic crisis management model. This is where we can use a lot of data to measure these things. So it starts with prediction. I didn't just write predict because this is a predict conference, but I also don't mean telling the future here. This is just the stage where you assess likelihood of issues and incidents across the company's verticals and you're never going to get full coverage here. It's just about trying to get as close as you can identifying the issues that are present. After that, then you have your planning stage. This is all about building action plans. What we see time and time again in Storyful is that the foundations are not strong, processes fail especially at times of crisis. So you need to go over these action plans, rehearse again and again and again. So then, ideally, that's the end of the model. We never have to deal with a crisis. Unfortunately, that's not the case in reality. Usually, after a crisis starts, we enter the control phase, which is the company reactive. So they've changed from passive to reactive. And this is about implementing those action plans and mitigating against long-term damage. Only then, when the crisis begins to slow down, can you assess the damage. And what you're trying to do here is we're working to return to norm. It's not often that you can actually work to return to a better stage than norm, but at least you've learned something along the way. Unlike many classic models, it finishes with the learning stage, which is the integrations of lessons you've learned into your action plans, further strengthening the processes and starting again. So how can we use data in this? These are just some ways we can use data, or many. We can perform an analysis of the landscape. That can be internal or external. We can look at consumer behavior, and we can assess the audience. Now for audience, here I pretty much mean the general public. So during, and this is my favorite part, during the actual crisis stage, everything becomes a little more real-time, and stress levels start to rise, and people want answers instantly. So really what you want here is a real-time monitoring and threat detection system and solid crisis reporting. What we often see as horrible is at this time of crisis, the highest levels, they don't know where to turn. Their classic channels of communication have broken, and they just need to find a signal. Then afterwards, you can do your brand perception assessments, where you measure, engage to see, has your reputation changed because of this crisis? And once more, you can look at the landscape, assess, and see if you need to make changes. So, Storyville, have you heard of Storyville? We're more classically known for the verification of social media, specifically to provide publishers with information and content that they can use in broadcasts or online publications. So as Paul mentioned, we kind of started during the Arab Spring where there was a dirt of content that came through social media, because people there didn't have any other avenues to share this content. And as that happened, publishers quickly found out they didn't know who to trust, they didn't know what content was real, and they didn't know whether they were allowed to use it. And honestly, those problems still exist in 2018, but we have worked with the biggest publishers in the world to ensure that they have verified content at hand, especially during the bigger breaking news stories, for example, this week, the earthquake in Indonesia. So this is kind of how we work. I had to get a Venn diagram in there as well, and we use unique data. Usually, when I say that, it is to whom I'm talking to, it is unique for you guys. It's probably not that unique, but we've built proprietary technology on top of that, and we always have a human layer, which means our data sets, they come to a human to analyze. Now, we obviously use a lot of machine learning to assist us to do that, but we still trust our team of analysts, strategists, and journalists as well. So how do we do this? You see some data inputs here on the right for today. I specifically focused on social media. This is not an exclusive list, but it does take in a large portion of social media space. For some of these platforms, we use API access. For some, we are a partner, and for some, we have elevated access as well. And I won't talk about this too much, but we suffer the same problems that I've heard in many other conversations this morning, which is we have to normalize this data. As you can imagine, the data from Facebook is very different from the data you get from 4chan. So we have to track the information flow, sequence this event, trace your origin to ensure that our data, the veracity of our data is in place, unify it, and sensorize it. And then, when we have all of that done, we work to understand the context of the data, separate fact from fiction, especially in the age of myths and disinformation, and work to verify authenticity and sources. So after that, then, this is our battle to create actionable insights. We have our inputs, which includes existing research, hypothesis, brand pillars. We look at trends from proprietary to digital. Social media monitoring tools, as well as our own data sets. And we try to blend in strategic vision and competitive intelligence. At that research phase, then, we try to attune as much as possible our insights towards that product or brand. We are usually working to test an assumption that a client has brought to us and then provide a layer of analysis on top of that that includes building customer journeys, brand touch points, understanding consumer motivations, and often very important is purchase considerations, as well. So very quickly, then, I'll just run you through a project that we're currently working through at Storyful, which is to answer those two phases of crisis management, control, and we've called it CEMIFOR, which is like an old-timey system of flag signaling. And really, what it's trying to do is not only track how a crisis spreads on social media, but also react to it and to give our clients enough time that they can make rational decisions. So this is kind of how a crisis spreads. Usually, and again, this is quite like a pathogen in its kind of how it spreads. So it usually starts with a post or a small group of posts. Let's take a product defect, for example. A picture can start on Instagram. From there, then, a contamination occurs where these original posts, they gain some traction through engagement on that platform. It spreads, then, through engagement, shares, retweets, and essentially, and a moment of real importance, is a shared cross-platform. So it's no longer contained on a single platform. Once more, it starts to bloom, where these two spreads pick the end. Usually, this is where mainstream media come in and kind of blow it out of digital media coverage. Makes everything exponentially larger. And then you reach your peak. Now, this is the point at which the crisis begins to slow or diminish due to company actions or the moving news cycle. So you can kind of see the exact... This is a real-life crisis. You can see the exact same thing happening here. What we've done with this project is we're trying to alert as close to origin as we can. So if you see the pink dotted line, that is the threshold point. So we've built a dynamic threshold for every single platform where if it goes above that threshold, that's the point it sends an alert to us. So to make this actionable and workable, our analysts are able to use flexible Boolean querying, which allows for a very broad query to be set up or a very specific query. We've also used historical backfill, which allows us to set those thresholds right there and then. And my favorite part, what we're working on right now is that these are smart alerts. So we're trying to separate, using machine learning, we're trying to separate out spam, non-entities, and make these alerts as smart as possible. And finally, clever integrations. This will only work if we can get the alerts to the person at the right time and actually, you know, wait them up if needs be. So we've actually integrated semaphore with email, Slack integrations, even phone calls if we need to do it. So right now this is a project that we're still working on. It's not out there in the market, but it's something hopefully one day we can work with a client on and hopefully true doing this will mitigate these reputational crises. I've been Dara Healy. Thanks very much. Enjoy your lunch and enjoy the rest of the conference. Thank you, Dara.