With traditional advanced analytics reporting, sending out such weighty tomes isn't all that unimaginable. In fact, Kimberly Holmes, head of strategic analytics at XL Group, a global insurance and reinsurance company headquartered in Dublin, Ireland, produced just such a document earlier this month. Such was what her team needed to do in order to present results of a large, univariate analysis involving close to 300 variables, she told me in a recent phone interview.
At 50,000 records, the size of the dataset involved in the project wasn't overly large. But her strategic analytics team was correlating about 270 variables against the experience in those 50,000 records, Holmes explained. "If we took a subset out, like for the 48 states we had in there, and looked at differences in experiences by state -- well, that means sitting down and comparing 48 exhibits."
Holmes, who has been leading XL's advanced analytics initiatives for two years, said she sees traditional reporting as sorely inadequate for conveying the information from her team's work. Selecting the information to include in a report even as sizable as 1,000 pages requires analysts to make judgment calls and, on the business end, "clearly understanding the patterns within 1,000 exhibits is impossible for the human eye."
XL is ready to reduce its reliance on traditional analytics reporting, as I discussed in yesterday's post, Insurer Sees Visual Analytics as Must-Have. Rather, it will let data visualizations -- à la SAS Visual Analytics (from this site's sponsor) -- tell the story. Holmes said:
We can put that information in Visual Analytics the next time -- which will be soon -- and the business will have a better understanding of it and even my team will have a better understanding of the story that's in the information. It's going to be a whole different experience going through the information with Visual Analytics on the screen and the business saying, 'Oh, isn't that interesting. Let's look at this.'
Presenting the data visually will lead to the business users raising questions they never would have understood to ask before as they stared at gray blobs of statistics. And the interactivity of Visual Analytics will let the analysts immediately show them correlations and patterns in the data based on the questions they ask, Holmes explained. "And all of this will result in better tools and better insights into what's driving risk and help us make better business decisions."
In the particular case of the univariate analysis recently completed, "having the results in Visual Analytics would have had a huge impact on the discussion with the business about what does and doesn't make sense," Holmes added.
In that case, we had to make judgment calls. We couldn't analyze all the things that looked to be powerful variables in terms of predicting experience for them. We couldn't say which two variables were saying almost the same thing as each other. We had to go back and do the next stage of the analysis. That changes with Visual Analytics. Because it has analytics capabilities, and is hooked up to the whole SAS arsenal, we could have answered a lot more of those questions as well in terms of why variables are showing up as powerful, what variables might be correlated with each other, and what's going on.
In fact, Holmes said, one of the first things her team will do once they get access to Visual Analytics (which it will use via a SAS hosting arrangement) is load up the dataset that supported the 1,000-page exhibit. It'll use that to build skillsets, do some training, and create demos to give to XL's leadership team. "They'll be very amazed by the capabilities."