Preventing Fire with Data
Nathaniel Lin, NFPA's new director of data strategy and analytics,
on how the power of big data can be harnessed to improve all manner of fire and life-safety initiatives. BY JESSE ROMAN
THESE DAYS, EVERY SECTOR of the economy is affected by the big-data revolution. Terms like “data analytics” are routinely used by professors in business school lectures and by day laborers calling into local sports talk radio shows. Netflix tracks subscribers’ viewing habits to inform the development of new shows. Cities around the country collect and analyze huge amounts of data to help drive policy and gain efficiencies. Data sets help fight the spread of malaria, grow better crops, and even identify ways to improve world happiness.
Closer to home, NFPA is looking to data analytics to help reduce fire deaths and property loss. In September, NFPA hired former IBM data scientist Nathaniel Lin as director of data strategy and analytics, yet another decisive step the organization has taken into the world of big-data analytics. Lin and his team will use the streams of data that NFPA and partner organizations have collected for decades, as well as new sources, to glean valuable insights and build data-driven tools and models to inform NFPA decisions and help stakeholders do their jobs more efficiently and effectively.
Lin thinks NFPA is sitting atop a “gold mine” of information; opportunities are everywhere, he said, from data-driven models that inform wildfire risk reduction strategies to analytics tools to help enforcers improve their inspection programs. “Data, with the right type of advanced analytics, is truly transformational,” said Lin, who has worked in business analytics for Fortune 500 companies including Fidelity Investments, AT&T, and IBM. “I felt that NFPA, with the mission to save lives and property, would be a very good use of the analytics skills I have accumulated.”
NFPA will work with a variety of organizations to develop big-data-based solutions for a host of challenges. More broadly, NFPA sees the effort as a catalyst to enable the entire fire and life safety community to join the big-data revolution. Organizing efforts are already underway, said Kathleen Almand, NFPA’s vice president of Research, who is overseeing the new initiative. “We hope to become the focal point for this activity within the community,” she said. “There is a lot going on out there with data, but it’s very disconnected. We have the national platform to serve the community around this. It is a natural fit for us.”
Since joining NFPA, Lin has worked to assemble the building blocks of the new initiative, including the development of a big-data strategy. New tools, platforms, and products based on NFPA’s data analytics research could be implemented and available for use as early as 2016, according to Almand.
Nathaniel Lin, Strategy & Data Analytics. Photograph: Adrienne Albrecht.
Lin, a native of Taiwan, is a graduate of the University of Birmingham in England and the Sloan School of Management at the Massachusetts Institute of Technology. He researched advanced radar technologies and superconducting sensors for the U.S. Air Force for 17 years before switching careers. After attending MIT, IBM tapped him to lead the company’s marketing analytics group in Asia-Pacific. He has also worked as an adjunct professor of business analytics at Boston College, Bentley University, and the Georgia Tech College of Management, and is the author of the 2015 book Applied Business Analytics—Integrating Business Process, Big Data, and Advanced Analytics.
NFPA Journal spoke with Lin about his new role at NFPA, what the fire and life safety world could gain by advanced data analytics, and how soon NFPA’s new data strategy could pay dividends for both the organization and its stakeholders.
Your nickname here is “the data monster.” How do you define your new role?
First of all, kudos to the senior management team for identifying this “data monster” as being an important part of NFPA’s strategy going forward. I see my job as leading the initiatives to use data to create value for NFPA—to increase NFPA’s value proposition, so to speak. We will be part of a data transforming team—“data monster” sounds to me like Godzilla going into town and smashing everything in its path. A transformer is different. A transformer comes in one form but adapts itself to different situations. If the situation calls for a jet, it becomes a jet. That’s how data is. Data, with the right type of advanced analytics, is truly transformational.
These terms get thrown around a lot, but for the record, how do you define “big data”?
That’s a very good question. Big data is really a misnomer, if you think about it in terms of volume. In big data, the “big” is defined in terms of big challenges and big opportunities. Big data can be characterized by four Vs. The first is volume—there’s a lot of it. The second is variety—data isn’t just numerical but exists in many forms. It can be text files, audio, or audio-visual, it can be telephone call center logs—all kinds of things. Within these varieties, some data is easily defined and stored because it is well structured. Some data is very unstructured. For example, if you take news clippings and digitize them, there is a whole bunch of text, but if you want to go back and ask how many of those newspaper clippings contain instances where a fire hose failed, you can’t do it without someone looking back though 70 years of data. Big data can provide structure to that unstructured data, and you can actually start doing something with it.
What are the last two Vs?
The third is velocity. During a fire, some of the data can be used in real time. One example is a biometric device like the Fitbit that checks pulse rates. Imagine this scenario: a firefighter’s baseline heart rate is 60, and when he goes into a fire his pulse rate jumps to 120 because his adrenaline is pumping. But then you see in real time that his pulse rate goes to 140, 150, 160 and climbing, because the person is dehydrated and he’s in distress. That data is transmitted in microseconds back out to the battalion commander’s tablet or mobile device. He can call on the radio and say ‘Hey Joe, get out of there,’ and you’ve just saved a firefighter’s life. So velocity is very important.
The fourth is veracity. The challenge is when you want to make something out of a mixed bag of data—where you don’t know much about it and there are lots of unknowns. Data that is very large, varied, and coming in fast and furious is a big challenge, but also a big opportunity. If you can integrate, manage, and analyze it, then you can use it to predict, perform, diagnose, and prescribe solutions. That is something I hope to bring to NFPA. In fact, some people add a fifth “V” for value. Properly applied, big data has been shown to generate big values in many applications.
The use of big data and data analytics seems to be nearly ubiquitous now across all sectors. Where is the fire and life safety community along this arc?
For fire, NFPA may be at the leading edge. Relative to all industries, though, the fire and life safety community is at the beginning. If you think about data analytics, it’s like a tsunami—it hits different industries at different times. Analytics started to hit in this country in the 1980s or early 1990s with catalog mailers applying analytics to predict what to put in the catalogs and who to mail them to. It moved to tech companies, hospitality, finance, telecommunications, and health care.
I would say the fire industry is almost at the tail end of that wave. But there is an advantage to being at the end because everyone else has already made their mistakes, and we can study the best practices that can make our adoption curve a lot shorter. If NFPA can adopt these best practices and leverage its position and current knowledge, assets, and relationships, instead of the process taking a decade we can do it in two or three years. To give you some context, when I first built my team at IBM in 2001, the analytics tool we used cost $1 million and the model building-process took a couple of months. The tool today is faster, much easier to use, and a lot more powerful—models can be built in days—and costs nothing. It’s freeware. So you can see how things have changed.
What areas at NFPA are ripe for further exploration in regards to data, and adding some of the value you spoke of?
Opportunities are everywhere. Public fire, wildfire, engineering, codes and standards, IT, business strategy, marketing, sales, research—all of these areas are ripe for the application of big data and analytics. Basically, wherever there is data in numbers and text, and wherever we need a solution or insights.
Can you cite a specific example of how this process of gathering and analyzing data might produce something meaningful for our stakeholders?
There are many communities in the wildland/urban interface (WUI) that are prone to fast-spreading wildfires. Say we take all the data pertaining to those communities—structural data, data from National Fire Incident Reporting System (NFIRS), news clippings—and screen for wildfire. All of that could give us useful data points that could be predictors of fires, such as locations of past fires, weather and drought patterns, road constrictions that make it difficult for firefighters to get in, distance to the nearest water source, types of building material, clearance around properties, and things like that. With big data we can potentially produce a model that could predict with some certainty the probability of a community or property experiencing a significant loss from a WUI fire in any given season. For instance, for dwellings in this particular community, with this particular roof, at this distance from the nearest home, this is our prediction of the probability of a house catching fire this summer. We can turn a lot of what was previously considered unusable data into something not only usable, but knowledge we can reliably use to plan fire protection strategies.
What would be the real-world application of that model?
We could potentially use that information to score communities on their wildfire probability—NFPA could say with some certainty that a community, by not doing mitigation, has increased the probability of a significant impact from wildfire from 50 percent to 80 percent. The community could also use our model to bring the fire probability down. We’re not just giving a blanket recommendation—we can also find the factors that will significantly reduce their risk at an acceptable cost. Data will train the model. We will be able to predict with some certainty different approaches to reduce the probability and severity of fire in a given community.
Are you saying you can predict the future with data?
Yes and no. We can’t use data analytics to predict what you will eat for breakfast tomorrow, but we can use it to predict, with high certainty, the probability what people like you will eat for breakfast. It’s grounded in reality. If I’d said the same thing 10 years ago it would’ve sounded like science fiction. Now it’s based in reality because I’ve seen it happen in other industries—the tools are in place. What is needed is the way to execute it.
What are your thoughts on the quantity and quality of data you have to work with at NFPA and in the fire protection world in general?
NFPA is really sitting on top of a gold mine, not only with data but also in its relationship with stakeholders. We have the data, both the publicly available data and the data we’ve gathered ourselves. We also potentially have access to much of the data in the hands of our stakeholders. The power of data analytics and big data lies in how well you can integrate it; the more you can integrate, the more opportunities you can derive out of it. This is why at NFPA we’re talking about how a 360-degree view of data is a lot more powerful and a lot more amenable for extracting significant insights.
What do you mean by a 360-degree view of data?
It means essentially that you consider all dimensions of the data, even when you think the data may not make sense. My mantra when I taught MBA students about applying data analytics was, “Always try it.” Just try it and let the data speak for itself. Nowadays the analytics tools are widely available— it isn’t difficult to find a powerful tool that can slice and dice and model the data quickly, and in so many different ways. That’s why the best way to get answers is often to just try it. Let the data speak.