Big data is big business. For the insurance industry, the volume of the information now stored in data warehouses is overwhelming. Don Canning, vice president of SunGard's insurance business, compares it to an ocean. And here, he explains how industry companies are beginning to use hundreds of billions of megabytes of data to make better assessments of their risk appetite than ever before.

When we talk about "Big Data" for the insurance industry, what exactly do we mean?

Don Canning: I was at Prudential Financial for seven years with my last, highest position being their chief technology officer for group insurance. But for the first three and a half, I was with the actuaries in the individual life business building data warehouses and "Big Data." This was a while ago but nevertheless, 15-to-20 terabyte warehouses back then were extremely enormous.

Some of the technologies we used are known as "massive parallelism," which is essentially the art of taking a database that has, lets say, a billion rows of information in it, and splitting it in half. Then you have two half-billion-row databases.

Let's say it takes 24 hours to read a billion rows and that's just unacceptable to an actuary. You can constantly break that data down into discrete, atomic components where this massively parallel architecture insulates the user from having to know where the data is and how to ask for it. The massive schema does that; the user just asks the question and the architecture figures out how to go get it.

So this was the first time you could, with an architecture, throw money at it and just get more hardware to get the return time down to six or eight minutes vs. the 24 hours. That's the first thing to note about Big Data in terms of taming the beast: massively parallelism is the secret architecture to addressing it. And many insurers are doing just that.

RM: So the volume of data we're dealing with today wouldn't have been practical to use 15 or 20 years ago even if you could have had it?

Canning: Exactly. That's the challenge. There is this thing that I call the "stream of consciousness" with an actuary, which is probably about 90 minutes. If you can't get them the answer to their question - their inspection, their analysis or their quantitative responses - then they're pretty much drifting off to the next subject. So what we strive to do, and what I did back when I was at Prudential, is to get that answer within that time frame.

Today, with SunGard's risk management architecture, we've really up ? ed the game over the past 15 years and, especially, over the last three. If you think about the actuarial sciences for risk management, you can think about it in two parts.

The first part is the engine: doing the calculations. It's what the actuaries use with rules-based, principlesbased or sto chas tic-type calculations to create the model of the [insurance] portfolio. Then you have what's called "stochastic on stochastic," which is to say "OK, this thread of volatility takes off in this direction, but what if I change it slightly along with an economic indicator? Or a catastrophe or a pandemic? What happens to my reserving levels then?"

When you have that kind of inquisitive mind sets that are trying to do full stress testing of the portfolios, you need a calculation environment that can do things like parallelism and high-performance-computing grids. So that is part A: being able to scale these equations to get them done very quickly.

The next part is that you really need a database - an operational data store - to begin to feed these engines at high speeds and receive the answer sets at high speeds. So that is kind of a "results database" that has to be finely tuned to pump that data into that engine as quickly as possible and get it back out of there as quickly as possible. You don't want to be bottle-necked by having to move data at that point. You want that calculation engine to be as crisp and as clean as possible.

Then, on the back end ofthat, once you say "OK, these results look good" where do you put it? That's the Big Data warehouse: 30 years of data. This is the warehouse off of which you build all your analytics and models. This is where the actuaries would spend a lot of their time drilling into.

RM: So how do companies actually put all this to use?

Canning: If you have a customer like MetLife, they're looking at hundreds of terabytes of data for patterns to gauge volatility. These are the capabilities that help develop new products, or tweak an existing product.

Ten years ago, actuaries were saddled with working into the wee hours of the morning trying to get the calculations done for the March close. Then, they became able to manage these calculation engines with these operational data stores to format the results to get them out. Now, that they have these enormous data warehouses, they are analyzing the data and looking to turn these data assets into capabilities that drive the business.

They can give better results on how well the carrier is doing in terms of meeting their risk appetite, how various products are performing, where the gaps are, what the trends are. The information is now moving from having an accounting, or retrospective, use to becoming proactive, forward-looking information. That is the exciting part of what Big Data is offering insurance companies.

What are the some of the other major applications of Big Data we can expect to see from insurance companies going forward?

Canning: What we've been talking about is just at the department level. If you start thinking about the fusion of a multi-line carrier or a multinational carrier, you're thinking about insurers like Travelers, The Hartford, AXA, ING, Zurich - these companies are buying other insurance companies around the world. They have to rationalize these product lines, they have to streamline all the supply chains, they have to get the branding levels right and ensure service levels. It's very complex.

Moreover, they need to understand their risks from a geopolitical and product perspective. So not only do you have all the aspects of the actuarial, departmental calculations, you have this information aggregation at the line-of-business level. And you have it at the corporate-enterprise level.

The corporate actuaries want to see the information in the aggregate but also drill down into each one of these pockets individually. When you have this kind of Big Data being formed, you are really able to see the risk appetite being integrated into the business.

What do you mean by integrated into the business?

Canning: There are three big levers in the insurance market. One lever is to grow the business organically: portfolio growth, new business, new products, new development. All these things. The middle lever is about operational efficiency: managing your costs, making sure the actuaries are reserving the right liquidity levels. The third lever is making sure those actuaries are giving the investment managers the capability to take all the cash and invest it.

When you think about these three levers, you're constantly tweaking them a little bit each day to match the risk appetite of the carrier. In order to do that, you have to bring these three pieces of the business together and the chief actuary is someone who is in a pivotal position to work with the chief risk officer, the CIO, the CEO and the chief investment officer to bring all these dimensions together to then balance that risk appetite.

Big Data is the vehicle that enables that capability.

These levers arent new and these adjustments are something that insurers have always been making. But now they can do it with more certainty and have more quantitative rationale backing their decisions?

Canning: Exactly. And then they can measure it. They can touch a lever and see how it affects results over the next 30 days. Then touch it again and measure that.

When you have Big Data - when you have billions of pieces of information - it's like an ocean. You begin to measure the rhythm of the ocean not the individual components of it. When you're looking at a billion of anything, you're really not looking at anything. You're observing it and looking for patterns.

When you have Big Data - when you have billions of pieces of information - it's like an ocean. You begin to measure the rhythm of the ocean not the individual components of it. You're really not looking at anything. You're observing it and looking for patterns.