Big Data-it’s a term that’s everywhere. It’s in newspaper headlines and in business journals. It’s in boardrooms and classrooms. It’s on cable news, network news and even in sports news. Five years ago, “Big Data” didn’t even exist, but today it seems to be in the middle of everything-the center of a phenomenon that’s sweeping across our social and economic landscape. Medicine, transportation, government services, and baseball, it seems like everyone is using Big Data to their advantage. And we, in the fire service, are no exception.
But we have to ask the obvious question: What is Big Data? What does the term even mean? Simply put, “Big Data” can be defined as any collection of data that is too large to be processed by any of the standard tools commonly used to work with data. In other words, it’s the new, popular name for the vast expanses of electronic data that are available today because of the revolutionary advances made in technology and telecommunications over the past decade or so.
These days, nearly all records are electronic, no matter what you’re looking for. Financial records, tax information, geographic information, even football stats are all electronic, very accessible, and full of valuable information-if you know how to use them.
In this article, we’ll provide an overview of the Big Data landscape for the fire service. We’ll talk about some of the more significant uses of Big Data and what we need to do to actually see them happen. Most of what we’ll discuss is not just theoretical; there are departments out there today that are already actually doing these things. By providing an introduction to the world of analytics, the hope is that more departments will soon follow suit. There are significant gains to be made for the fire service. We’ll provide a starting point and talk about what kind of Big Data is out there and what the fire service can do with it.
One has to wonder, after all the propaganda and hype, can this “Big Data” actually help us do our job-fight fires? The answer is yes, definitely yes, and the depth of the benefits might surprise you. Imagine being able to predict your workload for the coming tour so you’d have an idea of exactly how busy you could expect to be. Or take it a step further and imagine being able to predict where and when a fire was going to happen, not the general area but the actual addresses of the buildings with the highest risk of fire on a given day. Or imagine knowing exactly what factors, both avoidable and unavoidable, affect the response time of units in a given area and by how much.
What if we could put those factors on a map, showing the local streets as they actually appear on a given day? There would be undeniable benefits to that kind of ability. But these aren’t just hypothetical musings; there are fire departments that are actually already doing these things. How are they doing it? By using Big Data; and it’s just scratching the surface of what’s possible.
Making Data Big
Before we talk about the possibilities, we first have to talk about the data itself. After all, without the data, none of this is possible. So exactly what kind of data are we talking about? Who has it and how can we, the fire department, get it?
As the Big Data landscape continues to unfold, it’s becoming increasingly clear that no data is completely useless (so long as it’s accurate). But, there is certain data that would be especially valuable for a fire department (see chart below).
This list is by no means exhaustive; it’s only intended to be a sample of information that might be useful for a fire department to have. Any information that might help to describe your area (and the buildings in it) can be valuable. There might be a critical factor that’s unique to your area that’s not on the list; maybe “elevation” or “flood history” is important to your area. The possibilities are endless, and if you can measure it, you might be able to use it. But in reality, not everything is measured, not by the fire department at least. But in today’s world, chances are that somebody somewhere does actually measure it and they probably keep records somewhere too. Now, the trick is to find out where that info is and somehow get our hands on it.
Sharing With Our Friends
Despite how it may feel sometimes, the fire department is not the only entity working out in the field. Municipalities of every size have agencies that are responsible for all the things we don’t do. Depending on the size of your municipality, there may be quite a few of these agencies, and they might be doing work that matters to the fire department. Unfortunately, communication between these agencies is often lacking. Other times, it might not exist at all. The problem is that these other agencies may know something we don’t and, if we don’t communicate, we have no way of finding that out. In fact, they might actually have exactly the information, the data, we need. So how can we get it from them? We can share.
Data sharing is one of the most essential elements to the Big Data movement, and it’s especially important for fire departments. It would be impossible for us to somehow collect, organize, and maintain all the information that could be of use to us. The information that we do have (fire incident and emergency response history) is tremendously valuable, but it’s not enough. Building and construction permitting, zoning information, financial information of building owners, neighborhood demographics, criminal and complaint activity-these could all be just as valuable, if only we could collect it. Fortunately, we don’t have to, as there are others who are already doing it for us. All we have to do is get the info from them.
This is where we share. The fire department can share the data it has and get the data it needs in return. Keep in mind, many of these other government agencies lack the boots-on-the-ground presence that fire departments have, which makes us a valuable partner for them. They might not have the staffing to gather the data they need, and that might be the very data we have to share.
Data sharing can happen in several different ways, and it doesn’t always have to be complicated. Simply establishing and maintaining a regular means of communication between agencies might be enough. For example, a weekly e-mail of new building permit information, or a quarterly update of changes in property ownership information, could get the job done. For larger departments, it might actually be possible to create a fully automated process by which data is automatically sent from one agency to another through a sophisticated system. The key point is that the communication and data sharing are ongoing; we want to be sure we have the most current and accurate information to work with.
Another way data sharing can happen is by sharing access to a common database, with both agencies entering and using data relative to a shared venture. This could be a viable option for situations when another agency is responsible for maintaining a system firefighters would regularly use, such as fire hydrants or fire suppression systems. Firefighters could enter inspection or operational information they discover in the field, while the responsible agency enters its service and maintenance information. The important point here is that, by having a common database, both parties are getting information in real time, as it happens. For example, if you have a shared system concerning fire hydrant inspections, you can see exactly if and when a defective hydrant is repaired, as soon as it happens. By sharing data through a common system, both parties are better informed and provide a more effective service.
But who are we sharing with? Who exactly has all the building and geographic data we can use? The answer to that depends on where you work. Each municipality is set up differently, but some key responsibilities will certainly be handled by someone. You’ll have to find out who does what in your city. But generally, you should start with the following:
- Agency that regulates building construction and use (Department of Buildings?).
- Agency that oversees land use and zoning (Department of City Planning?).
- Agency that oversees property ownership and taxes (Department of Finance?).
- Agency that handles nonemergency complaints (Police Department?).
Putting It to Work
By now, we know exactly what kind of data is important, where we can find it, and how we can get it. We have Big Data! That’s great, but … what do we do now? After all, having the data doesn’t mean anything unless we do something with it.
If we were all scientists and this was a research journal, we might actually care about the academics of it all. But we’re firefighters-we don’t do math for fun. We just use it when it helps us fight fires. So, how can we use all this data to help us in the field? The possibilities are too many to list here, but the following applications are a good place to start.
No one knows where and when a fire will happen, but we can guess. Often senior firefighters can guess with startling accuracy. They may call it just a “hunch,” but in reality their guess is based on an assessment of their past experience applied to present conditions. They may know that when the weather turns cold, there always seems to be a job in that old row of tenement buildings around the corner. And, more often than you’d expect, they may be right. Wouldn’t it be great if we could all have a “hunch” like that?
With Big Data in our corner, we can. We may not all have decades of quality fire experience, but we can all have decades worth of quality fire data, which is enough to make a pretty good guess of our own. An amazingly good guess, in fact.
Through rigorous analysis of a variety of data, it’s possible to calculate the relative likelihood that a building will have a fire incident by looking to see what, if anything, is correlated with the incidence of fire. What actions or conditions are statistically significant in predicting a fire incident in a given building? Is a nonfireproof tenement more likely to burn than a commercial occupancy? Does a history of building code violations make a fire incident more likely? What about a history of noise complaints or rodent complaints? Listen to what the data tells you-you might find some unexpected, but useful, connections.
The result of all the analysis is that every building in an area can be assigned a level of fire risk. The more rigorous the analysis, the more specific the risk classification can be. A basic analysis can provide a division into broad categories, such as “high risk” or “low risk,” or a more detailed analysis could yield a more specific, numerical “risk score,” which could be used to create a more intricate ranking system. In the end, you have a list of the buildings with the highest risk of fire in a given area.
But fires aren’t the only thing Big Data can predict. We respond to much more than just fires: Medical emergencies, natural gas leaks, stuck elevators, cats in trees (does that really happen?)-you name it, and we respond. That also means we have data on it. Throw that service history data into the Big Data universe, and we can make some powerful predictions about the workload for a given tour.
The true power of these predictions-the ability to forecast workload and fire incidence-lies in what we can do with them. Information like this could easily prove vital to chronic issues in resource allocation for a department of any size. It could be used to inform staffing increases or redistribution to meet projected demand in certain areas or at certain times. Fire prevention resources could also be more effectively allocated by prioritizing inspections in higher risk buildings.
Beyond resource allocation, fire forecast information could prove beneficial to field units as well. Situational awareness is a critical concern to responders in the field, and workload and fire forecasting can be a tremendous asset in that regard. By being aware of what to expect, and where to expect it, firefighters will be better prepared for whatever awaits them during the upcoming tour.
For many years, any discussion of resource allocation or performance metrics in the fire service ends up in the same place: response times. The widely accepted opinion is that the effectiveness of emergency response can be accurately measured by response times. The truth, as many of us are aware, is much more complex than that.
Response times are always the principle metric at fire department staff meetings, budget hearings, and community meetings. This truism means that you can count on consultant groups to focus their analyses on these metrics to determine your resource allocation and optimization. Moreover, in some consultancies, this is the only metric used to develop resource optimization recommendations.
Now that there are myriad metrics accumulated at the municipal level to incorporate into risk and vulnerability scoring, response time becomes but one piece of the equation of optimally deploying or relocating resources. Response time of the initial unit to arrive was often viewed as the conduit metric between the caller and fire department intervention; but this has become cloudy, as any fire department can experientially tell you that measurable response time should also involve the arrival of the entire complement of units dispatched for a particular incident. With this admission from fire department planners, and realization by the academic community, there are currently several academic projects underway attempting to determine predictive algorithms to quantify the effectiveness of both initial and complete response times. With Big Data at our fingertips, we are finding the needed help to bring a little more clarity to the response time issue.
Considering the scope of this article, we will not get into a deep discussion of the complexities of response times. Rather we’ll focus on the role Big Data can play. Through data analysis, we can paint a more complete picture of the response time process and possibly answer some very important questions that have long surrounded the response time conversation. What other factors influence response times? Are there local conditions that have a significant effect on response times, such as traffic patterns, street conditions, population density, or any of an array of other possibilities? How does a school affect response times? How about speed bumps? Or a street riddled with potholes? The questions go on and on.
In the end, proper analysis can give us a list of factors that influence response times, along with the relative strength of each factor. For example, we can figure out the numerical impact of a street full of potholes or exactly how much time each speed bump costs us, depending on the time of day. This information can be of tremendous use in explaining the actual performance trends of fire companies, especially when political pressure turns its attention to the threat of firehouse closures.
Gone but Not Forgotten
For many of you, this may sound familiar: Using data and statistics to “improve” fire department performance through “effective resource allocation.” It sounds familiar because it’s been done before-more than 30 years ago-although it may be more accurate to say it was attempted before. The results back then weren’t exactly what was hoped for.
In the late 1970s, the RAND corporation conducted an in-depth study of fire department operations in New York City, in the hopes of combating the growing fire epidemic from the economic hardships the city was facing. The idea was to use data-based analysis to make recommendations to the FDNY to make the department more efficient and effective. The result was the advent of a new metric of performance-response times. Response times were recorded and analyzed to the furthest extent the technology at the time would permit. Despite the good intentions, however, the results were disastrous. The analysis suffered from several fatal flaws, which led to the recommended closure of several of the busiest fire companies in the most badly fire-ravaged parts of the city. To oversimplify a complex situation, the recommendations did not make anything better.
So if things went so badly last time, why should we try it again? Is the situation so different today that we actually believe the results would be any better for the fire service? The answer is yes, and for two general reasons: (1) Advanced technology has changed the situation, and (2) we have history on our side.
The story of the NYC RAND project has been studied at length in the decades since it ended, and the lessons have been well documented. Perhaps the most comprehensive documentation can be found in Joe Flood’s book The Fires, in which he examines at length the circumstances surrounding the project and the shortcomings of the RAND process. Specifically, Flood identifies four key areas that led to the unintended consequences: (1) using faulty data, (2) using flawed assumptions, (3) the information and cultural gap between researchers and firefighters in the field, and (4) political influence.
By keeping these four factors in mind, we can learn from past mistakes and not allow history to repeat itself. But are we capable of actually doing that? We are. And what gives us such ability? Technology does.
Technology is the big difference. Over the decades since the RAND study, the unimaginable advances in technology have changed everything. And during that time, the pace has only quickened; the changes we’ve seen in the past 10 years alone have redefined what is possible. Which means that today’s technology lets us do things that we had not even imagined in the days of NYC RAND. So when it comes to data analysis, we can simply do it better.
First, there is the data itself. There is a saying in analytics: “Garbage in, garbage out.” That means that if your data is bad, your analysis will also be bad. Thirty years ago, response times were recorded by stopwatch. The officer would (in theory) stop the watch when they arrived on scene. The results were unreliable at best, but they were treated as facts. Today, we have the ability to automatically and electronically record data of all kinds, far beyond response times. Modern technology not only makes it possible to gather such information but also ensures its reliability to a comforting degree. It gives us more data, and the data is good, thereby minimizing the influence of faulty data that plagued the RAND project. Good data means better analysis. Additionally, there is the speed and complexity of analysis that can be readily done today. Complex calculations take mere seconds today, making multiple iterations and redundant proofs practical, thereby increasing the accuracy and reliability of the analysis itself.
But the most significant change is not actually what we can do but rather who can do it. In the past, data analysis could only be done by the professionals (statisticians, mathematicians, and computer engineers). Today, technology has brought that ability to the masses. Software capable of complex analysis comes standard on your laptop. This means that it’s no longer reserved for the academics who have no understanding of the reality of emergency response operations. It also means there is no longer the need to rely on private consultants whose recommendations can be far too susceptible to political influence (thus addressing the second of RAND’s four shortcomings).
Instead, those of us with real firsthand response experience can be involved in the analysis ourselves. Considering the accessibility and user-friendliness of today’s technology, firefighters themselves can more easily become educated in how to conduct analysis and make recommendations that actually make sense and ultimately help us do our job. And firefighters can learn data analysis far more readily than an analyst can learn how a firehouse works. As a result, the final two problems that most plagued the RAND study, flawed assumptions and the information gap with the firehouse, are problems no more.
The Future of Everything
Big Data is the next big thing. This is a fact that has not been lost on the fire service; already we have departments out there are putting Big Data to work. From forecasting workload to predicting fire incidence to identifying factors that influence response times, data analytics is already effectively being used in our field-and it’s only the beginning. The future is even more promising. What about the way we use geographic data and the potential for mapping or even geolocating missing members? Or perhaps the potential to explore the intricacies of the way we operate through analysis of fireground portable radio recordings? The possibilities are limitless.
But as we move forward, it’s important to remember that, while the possibilities may be limitless, the analysis itself is not. There are definite bounds to the science of data analysis and, while it may be an extremely valuable asset, it is by no means the magic solution to every one of our problems. Big Data can offer incredible insight, but its benefits should be taken with a grain of salt, as no data can ever truly tell us the full story. After all, statistics are not sorcery and, like anything else, have limits.
The conclusions we make from our analysis are only as good as the quality of our data and the validity of our assumptions. It is critical that we keep this reality in mind as we move deeper into the world of Big Data, lest we run the risk of repeating the mistakes others have made before us. But if we carefully bear these limitations in mind as we leverage the technology and resources at our fingertips, we will continue pushing the boundaries of possibility, and the results will be exciting to see.
Whether we’re ready for it or not, we live in the Age of Information, an age where technology is everywhere and information flows freely to those who are willing to look for it. We in the fire service, along with the global community at large, are just starting to capitalize on this information-this Big Data-and use it to make the world a better place.