If you're an actuary, you probably love data. Or studying. Or days off for studying. Or some of the above. And whether you love data or not, you work with it daily and most likely perform analyses or create models. As technology becomes smarter and faster, companies are finding more ways to track information and collect data. This influx of additional data is great for building models and testing assumptions, but can at times make other analyses more complex and complicated.
Excel is an actuary's best friend, but sometimes even excel can't handle the data if there's too much of it (I used to get the "not enough memory" error way too many times)! There are also quite a few limitations if you are only using excel to review and analyze your data. Based on my experience:
- Pivot Tables and Conditional Formatting are okay for reviewing data, but you can't compare multiple dimensions at once.
- Filtering and flagging data only goes so far - in many cases you are still scrolling through data reviewing other attributes.
- You don't get a good overview of your data in excel - it's important to know your population and it's not easy to get an all encompassing overview.
So let’s talk about data visualization. Sure, data visualization may not be necessary in some cases, but I hope by the end of this post, if you’re someone who thinks “why visualize data when I can easily perform my analysis in excel” that you may have more of an interest on how data visualization can enhance your work. Or better yet, maybe you'll get inspired and develop a proof of concept to show your boss in order to bring in a data visualization tool! Keep reading for five examples of how data visualization can enhance your work as an actuary!
1. Quickly Draw Insights Through Visualizations
No matter what type of work you do as an actuary, there may be times where you will need to review data quality to make sure the data is expected. Or maybe you need to check the outcome of a model to see what the data looks like. Typically, a lot of this review is done in excel and results in scrolling through rows and rows of data or analyzing pivot tables. There's only so much you can see in a table - this is why when you're reviewing loss development triangles, it's usually best to use the red/yellow/green conditional formatting. Using a visualization tool, such as Tableau or Power BI can help you quickly draw insights and identify the data points you should focus on.
Example: Suppose you need to review claim amounts and premium amounts for a specific period for personal auto policies. Maybe you also care about the deductible amount and age of the policyholders. You'd most likely use a pivot table to review this information, but you could miss the big picture. If you use a visualization tool, you'd be able to plot a scatterplot with claim amount and premium amount on the axis, with each point's size based on the deductible amount and color based on age or an age bucket.
One of the main reasons I love visualizing data is because of the versatility of reviewing multiple dimensions of data at once. It can save a lot of time and help you focus on the most important information!
Example: Suppose you need to review claim amounts and premium amounts for a specific period for personal auto policies. Maybe you also care about the deductible amount and age of the policyholders. You'd most likely use a pivot table to review this information, but you could miss the big picture. If you use a visualization tool, you'd be able to plot a scatterplot with claim amount and premium amount on the axis, with each point's size based on the deductible amount and color based on age or an age bucket.
One of the main reasons I love visualizing data is because of the versatility of reviewing multiple dimensions of data at once. It can save a lot of time and help you focus on the most important information!
2. Explore Characteristics of Attributes More Efficiently
If you're exploring attributes affecting a pricing or reserving model or doing data quality on specific sets of data, it may be more efficient to use data visualizations to being your review process so you know which characteristics of the data to focus on.
Example: You may be in a situation where you need to analyze claims for a specific line of business and find patterns for specific attributes of data. Are there specific zip codes that are typically reporting more damage on vehicles? Could the age or number of drivers on a policy be impacting this uptick? A visualization tool can help you explore multiple characteristics on one dashboard so you can see how they interact with one another. Once you know which characteristics you want to focus on, you can finalize your dashboard and any time you have new data you want to review, it's just a quick upload with minimal set-up.
Example: You may be in a situation where you need to analyze claims for a specific line of business and find patterns for specific attributes of data. Are there specific zip codes that are typically reporting more damage on vehicles? Could the age or number of drivers on a policy be impacting this uptick? A visualization tool can help you explore multiple characteristics on one dashboard so you can see how they interact with one another. Once you know which characteristics you want to focus on, you can finalize your dashboard and any time you have new data you want to review, it's just a quick upload with minimal set-up.
3. Visualizations Provide Better Ways to Filter and Compare Data
Data visualizations allow for a lot of versatility when it comes to comparing data, especially year over year comparisons. If you're doing data quality, you may have one set threshold for all of the data when it may make more sense looking at different thresholds for each state. This would add a lot more time to a review in excel, but using data visualization, these thresholds, when used with filters, would make the process more efficient and accurate.
It's also really easy to add in new calculations in Tableau or Power BI once the source data has been added - so if you're in the middle of creating a dashboard and want to see a new combination of data or bucket the data, you can easily do so in the visualization software.
Example: There may be a time where you need to compare data year over year to make sure different groupings of data remain consistent. Maybe you are looking at the loss ratio or a variety of other ratios. Depending on your dataset, there are a variety of ways you could visualize year over year data to compare, from bar charts to scatterplots to heatmaps. You could even use side by side box and whisker plots to see how the population of data has changed year over year.
When using data visualizations to filter and compare data, the possibilities are endless, as well as the insights that can be drawn from the data. Data visualizations can also make your comparisons and data checks more effective! Just another reason I think data visualization is a great alternative!
It's also really easy to add in new calculations in Tableau or Power BI once the source data has been added - so if you're in the middle of creating a dashboard and want to see a new combination of data or bucket the data, you can easily do so in the visualization software.
Example: There may be a time where you need to compare data year over year to make sure different groupings of data remain consistent. Maybe you are looking at the loss ratio or a variety of other ratios. Depending on your dataset, there are a variety of ways you could visualize year over year data to compare, from bar charts to scatterplots to heatmaps. You could even use side by side box and whisker plots to see how the population of data has changed year over year.
When using data visualizations to filter and compare data, the possibilities are endless, as well as the insights that can be drawn from the data. Data visualizations can also make your comparisons and data checks more effective! Just another reason I think data visualization is a great alternative!
4. Identify How Subsets of Data Affect Overall Population
There may be some situations where you want to see what your data looks like without certain subsets of data. There isn't an easy way to do this in excel. You could create a VBA macro that could exclude certain records with a click of a button, but why go through all of that trouble when you could do the same thing using visualization software and get a better overview of how different metrics change when excluding data?
Example: If you are interested in how specific lines of business are trending over time and you want to identify and exclude outliers causing large shifts, you could easily do this using visualization tools. With all of the additional features in tableau, you can also have visuals interact where you'd be able to hover over a bar in one chart which could highlight all of the data making up that specific bar in a different chart.
Example: If you are interested in how specific lines of business are trending over time and you want to identify and exclude outliers causing large shifts, you could easily do this using visualization tools. With all of the additional features in tableau, you can also have visuals interact where you'd be able to hover over a bar in one chart which could highlight all of the data making up that specific bar in a different chart.
5. Use Interactive Dashboards to Drill-Down to Reduce Noise
Depending on the work you do, you may find a lot of relationships in your data which you must use to make assumptions, create models or test for data quality. There may also be processes in place for you to adequately review these relationships. However, when done the right way, data visualizations and interactions between visuals can show you additional information in seconds that may take hours to find in excel. Interactive dashboards can also help you get to the problem areas of your data and see the makeup of those problem areas with quick drill-down features.
Example: When reviewing certain relationships, such as a loss ratio, you may see that a specific state stands out and you want to drill down to find the line of business causing the high loss ratio. You also want to see if it is one policy driving the loss ratio or many and if they have any other similar characteristics. In excel, you might be able to set specific thresholds and filter on these thresholds to find that data, but you could be missing some other important information. Maybe there is an issue with a specific product across the board, but you only found one point of data above the set threshold, so you let it go. Using a dashboard with multiple visualizations, allowing you to drill-down to the data could quickly alert you to these patterns without missing any specifics due to set thresholds.
I've found the drill-down capabilities of Tableau and Power BI to be very useful in identifying specific areas of data to focus on. It is a lot quicker to find these data points using a variety of visuals than it is to find these relationships in excel.
Example: When reviewing certain relationships, such as a loss ratio, you may see that a specific state stands out and you want to drill down to find the line of business causing the high loss ratio. You also want to see if it is one policy driving the loss ratio or many and if they have any other similar characteristics. In excel, you might be able to set specific thresholds and filter on these thresholds to find that data, but you could be missing some other important information. Maybe there is an issue with a specific product across the board, but you only found one point of data above the set threshold, so you let it go. Using a dashboard with multiple visualizations, allowing you to drill-down to the data could quickly alert you to these patterns without missing any specifics due to set thresholds.
I've found the drill-down capabilities of Tableau and Power BI to be very useful in identifying specific areas of data to focus on. It is a lot quicker to find these data points using a variety of visuals than it is to find these relationships in excel.
I have experience using both Tableau and Microsoft Power BI, but there are also plenty of other data visualization tools out there, too. Check out the articles below for more information!
If you are interested in applying data visualizations to your actuarial work, I'd be happy to speak more - just send me a message on LinkedIn use the contact form on this site!
If you are interested in applying data visualizations to your actuarial work, I'd be happy to speak more - just send me a message on LinkedIn use the contact form on this site!
RSS Feed
