In this blog post, Im delighted to refer readers to an interesting, useful new article from Aons flagship journal of ideas, One, on the topic of big data. In my earlier post on big data, I referred to a well-regarded report from the McKinsey Global Institute; although the McKinsey report is an excellent resource for anyone wanting to get up to speed on big data and how it is transforming virtually every sector of the world economy, I also noted that it says little directly about the impact of big data on organizational risk management and insurance programs.
The big data article in One, by contrast, is very much risk-and-insurance-focused. If youre like many of the risk managers I frequently speak with, you may have a pretty good idea of whats meant by the concept and early practice of big data management; however, you may be a little fuzzy on what it all means for you and the discipline of risk management (let alone how to explain to your organizations senior management how big data impacts any risk data strategies your company has already employed).
Almost anything you read about big data will, at some point, introduce some staggering, difficult-to-contextualize facts and figures about the exponential growth in the amount of data generated and captured nowadays. One tells us, for instance, that:
Data scientists talk today in exotic figures such as zettabytes, 42 of which could transmit every word ever spoken by humans in history. Global data, by one measure, equaled 2.7 zettabytes in 2012. Shockingly, that total was up 48 percent from just the year before. And it will only continue to grow.
What's so useful about the One article, however, is the way in which it contextualizes todays new data realities:
Nearly everyone working today is creating data as part of his or her job, from the CEO in the corner office to coffee-shop road warriors and delivery truck drivers, all the way down to multitudes of robots and sensors on high-tech manufacturing floors. Big data simply means recognizing that fact and doing something about it. [emphasis added]
When I was reading the One article, it occurred to me that many risk managers may already be well situated to begin doing something about big data; I'm referring specifically to those risk managers who have embraced the business value in analytics and reporting enabled by today's risk management software.
Consider this quote in the One article from information-management strategist Nigel Turner:
Coca-Cola has 57 million followers on Facebook. But there's an awful lot of noise in that. And it has to be exploited for a purpose. The mistake organizations make is to start collecting it and even paying for it, but you have to have a purpose in looking at it. And then you have the challenge in integrating it into your existing data. Otherwise you get swamped by it. [emphasis added]
I submit that risk managers who make effective use of risk management information systems have already gone a long way toward developing the habits of mind necessary to make a useful contribution to their organizations big data efforts. Risk managers are, for example, already collecting extensive claims and loss data, and their reasons for doing so are already well established: To identify trends that can be addressed by loss-control initiatives and/or potentially costly exposures that should be addressed by using traditional risk control techniques.
Admittedly, getting to the point where a risk manager has the basic control over information to enable a renewal process they are happy with takes time and effort. Yet, the experience gained from getting to that point ought to be seen as invaluable preparation for the much bigger (yet still parallel) challenges associated with big data projects.
The challenge of integrating new data into your existing data is where most risk managers begin with a big data project. As a result of hard-won experience, risk managers nowadays may take it for granted that their risk management information systems will be able to collect and integrate data from multiple internal systems (from human resources to financial to customer relationship management) and disparate external sources (including third-party administrator and carrier systems). Most often, the results of this integration will include optimized insurance program structures and coststhat is, the kind of outcomes that have demonstrable business value.
The One article concludes with some important observations by Neil Harrison of Aon Global Risk Consulting:
Data has always been important in the underwriting process, but now the depth and breadth of available data has led to a dramatic increase in focus. The quality of data can be a key factor in securing coverage, in obtaining the best possible terms and conditions, and in setting equitable premiums. On behalf of our clients, we collect and collate data that we then present to insurance markets and use as a negotiating tool.
Neil's point is underscored in numerous case studies that focus on how our clients are utilizing their data to ensure a complete understanding of the elements of this program in the marketplace. We say that "underwriters punish uncertainly" with higher premiums. eSolutions clients use data to eliminate this uncertainty and therefore ensure that their programs are accurately quoted. (See the Mosaic Company, Bon Secours Health System, and Parmalat for more on this.)
The payoff from utilizing data to gain an underwriting advantage can be huge. Now, I don't want to appear to minimize the challenges inherent in making big data work for your organization; yet, for the risk manager willing to explore the big data movement, there are even bigger prizes in store.
Ah, two buzzwords and one sentence: big data and predictive analytics. Essentially, these two are connected in that after you pull together your big data elements, applying predictive practices gives you a better understanding of your program and a better understanding of its direction. The emerging practice of predictive analytics offers additional parallels for the risk manager looking for structured approaches to big data.
At its core, the process of bringing predictive analytics into real-world use involves multiple components, including databases, statisticians, consultants, actuaries and data modeling experts along with in-house resources like risk managers.
And yet, the foundations of a successful predictive analytics project are similar to those of the three initiatives we've just discussed:
Whatever the effort, if it's data-driven, the key first steps will be fundamentally similar: Those steps will come back to thinking about the basic data elements necessary to help you get the new perspective you want out of the data you can access. Indeed, without a solid understanding of what your organization wishes to accomplish with its data efforts (whether big, medium or small), the pursuit could be fruitless. Start with a hypothesis of what you believe you will find and be prepared to be surprised.
To wrap up, I encourage you to read the One article, and as you do, think about the array of data coursing throughout your organization and how, if it were pulled together with the data you already control, what you could do with that kind of information?