I’ve lately seen some bad takes on data analytics in emergency management. For those not completely familiar, data analytics is a broad-based term applied to all manner of data organization, manipulation, and modeling to bring out the most valuable perspectives, insights, and conclusions which can better inform decision-making. Obviously, this can be something quite useful within emergency management.
Before we can even jump into the analysis of data, however, we need to identify the metrics we need. This is driven by decision-making, as stated above, but also by operational need, measurement of progress, and reporting to various audiences, which our own common operating picture, to elected officials, to the public. In identifying what we are measuring, we should regularly assess who the audience is for that information and why the information is needed.
Once we’ve identified the metrics, we need to further explore the intended use and the audience, as that influences what types of analysis must be performed with the metrics and how the resultant information will be displayed and communicated.
I read an article recently from someone who made themselves out to be the savior of a state emergency operations center (EOC) by simply collecting some raw data and putting it into a spreadsheet. While this is the precursor of pretty much all data analysis, I’d argue that the simple identification and listing of raw data is not analytics. It’s what I’ve come to call ‘superficial’ data, or what someone on Twitter recently remarked to me as ‘vanity metrics’. Examples: number of people sheltered, number of customers with utility outages, number of people trained, number of plans developed.
We see a lot of these kinds of data in FEMA’s annual National Preparedness Report and the Emergency Management Performance Grant (EMPG) ‘Return on Investment’ report generated by IAEM and NEMA. These reports provide figures on dollars spent on certain activities, assign numerical values to priorities, and state how much of a certain activity was accomplished within a time period (i.e. x number of exercises were conducted over the past year). While there is a place for this data, I’m always left asking ‘so what?’ after seeing these reports. What does that data actually mean? They simply provide a snapshot in time of mostly raw data, which isn’t very analytical or insightful. It’s certainly not something I’d use for decision-making. Both of these reports are released annually, giving no excuse to not provide some trends and comparative analysis over time, much less geography. Though even in the snapshot-of-time type of report, there can be a lot more analysis conducted that simply isn’t done.
The information we report should provide us with some kind of insight beyond the raw data. Remember the definition I provided in the first paragraph… it should support decision-making. This can be for the public, the operational level, or the executive level. Yes, there are some who simply want ‘information’ and that has its place, especially where political influence is concerned.
There are several types of data analytics, each suitable for examining certain types of data. What we use can also depend on our data being categorical (i.e. we can organize our data into topical ‘buckets’) or quantitative. Some data sets can be both categorical and quantitative. Some analysis examines a single set of data, while other types support comparative analysis between multiple sets of data. Data analytics can be as simple as common statistical analysis, such as range, mean, median, mode, and standard deviation; while more complex data analysis may use multiple steps and various formulas to identify things like patterns and correlation. Data visualization is then how we display and communicate that information, through charts, graphs, geographic information systems (GIS), or even infographics. Data visualization can be as important as the analysis itself, as this is how you are conveying what you have found.
Metrics and analytics can and should be used in all phases of emergency management. It’s also something that is best planned, which establishes consistency and your ability to efficiently engage in the activity. Your considerations for metrics to track and analyze, depending on the situation, may include:
- Changes over time
- Use of trend lines and moving averages may also be useful here
- Cost, resources committed, resources expended, status of infrastructure, and measurable progress or effectiveness can all be important considerations
- Demographics of data, which can be of populations or other distinctive features
- Inclusion of capacities, such as with shelter data
- Comparisons of multiple variables in examining influencing factors (i.e. loss of power influences the number of people in shelters)
- Regression modeling, a more advanced application of analytics, can help identify what factors actually do have a correlation and what the impact of that relationship is.
- Predictive analytics help us draw conclusions based on trends and/or historical data
- This is a rabbit you can chase for a while, though you need to ensure your assumptions are correct. An example here: a hazard of certain intensity occurring in a certain location can expect certain impacts (which is much of what we do in hazard mitigation planning). But carry that further. Based on those impacts, we can estimate the capabilities and capacities that are needed to respond and protect the population, and the logistics needed to support those capabilities.
- Consider that practically any data that is location-bound can and should be supported with GIS. It’s an incredible tool for not only visualization but analysis as well.
- Data analytics in AARs can also be very insightful.
As I mentioned, preparing for data analysis is important, especially in response. Every plan should identify the critical metrics to be tracked. While many are intuitive, there is a trove of Essential Elements of Information (EEI) provided in FEMA’s Community Lifelines toolkit. How you will analyze the metrics will be driven by what information you ultimately are seeking to report. What should always go along with data analytics is some kind of narrative not only explaining and contextualizing what is being shown, but also making some inference from it (i.e. what does it mean, especially to the intended audience).
I’m not expecting that everyone can do these types of analysis. I completed a college certificate program in data analytics last year and it’s still challenging to determine the best types of analysis to use for what I want to accomplish, as well as the various formulas associated with things like regression models. Excel has a lot of built-in functionality for data analytics and there are plenty of templates and tutorials available online. It may be useful for select EOC staff as well as certain steady-state staff to get some training in analytics. Overall, think of the variables which can be measured: people, cost, status of infrastructure, resources… And think about what you want to see from that data now, historically, and predicted into the future. What relationships might different variables have that can make data even more meaningful. What do we need to know to better support decisions?
Analytics can be complex. It will take deliberate effort to identify needs, establish standards, and be prepared to conduct the analytics when needed.
How have you used data analytics in emergency management? What do you report? What decisions do your analytics support? What audiences receive that information and what can they do with it?
© 2021 Tim Riecker, CEDP