Game Plan

While some burning issues in insurance technology become yesterday's news after a time, data management is always news. The fundamental challenges of data management remain, particularly data quality, and have grown more complex for a couple of reasons: the recognition of the need to maintain control of the input paths of data as well as the increasing realization of the value of unstructured data. This new awareness has driven more focus on data management strategies.

Larry Danielson, a partner in the consulting practice at Deloitte, doesn't believe finding the right data management strategy is a difficult task for carriers as long as they bring together people with the right expertise and business knowledge. "You have to get definitions, set up a governance structure, and then build it," he says. "It takes a good bit of work before you actually can see a business result. In these times, it will be even harder to justify [the expense]."

The mission of the insight and innovation team within Farmers Insurance is to promote science-driven decision-making, according to Murli Buluswar, vice president. "The idea behind insight and innovation was we thought there was an opportunity to use data more effectively to help us make better decisions," he says.

Buluswar has been with Farmers for just short of two years, and in that time he has brought in people with varying backgrounds. "I'm less interested in their credentials and more interested in their critical reason skills," he says. "Bear in mind, building the team took a while because you have to recruit the right set of skills–people who understand analytics but are not drowning in it to the point where they are unable to see the bigger strategy implications. Getting our arms around the data issues was a nontrivial task. We feel more comfortable about where we are today, but it has been a cumbersome process."

A little more than half of what his team does involves questions the members ask proactively and try to answer for the business units. The first stage in this process is the problem definition, explains Buluswar. Other steps include data gathering, statistical modeling, business analysis, and communication to the business unit in language that is relevant to it. Finally, the team partners with the business unit in an advisory capacity as efforts get rolled out. The Farmers team is bolstered by the software provided by SAS.

Buluswar's goal is to drive toward optimization. "It has a more clear-cut meaning when you are thinking about advertising efficiency–where do I allocate my advertising dollars and how do I get the maximum bang for it?" he says. "But the concept is similar in every function."

When Brotherhood Mutual Insurance moved from its Excel-based information systems, the carrier hoped to create an organizationwide reporting foundation through the implementation of WebFocus, business intelligence technology from Information Builders.

The first data project Brotherhood undertook involved the carrier's customer service department, according to Rob Fosnaugh, senior programmer analyst with Brotherhood. In addition to doing internal work, the carrier's CSRs would service customers for its independent agencies. "We needed to track both internally and against the agencies," he says.

The CSRs had a spreadsheet for internal work and for each agency they would work on, relates Fosnaugh. At the end of the month, there was a huge process of merging 40 Excel sheets into one to get the final numbers. Fosnaugh created a Lotus Notes application that allows them to gather all the data, and every day they can feed it into the carrier's SQL server. "On a daily basis, [users] get all the information they need, there is no waiting, and there is no manual intervention," he says. "Previously, it took 40 to 50 hours a month to get all the information gathered together to get the reports to show the president and the agencies. Using WebFocus, you have graphs, charts, drill-through reports. All the what-if questions are answered."

From there, Brotherhood moved to the claims area, where the company had been unable to see what the workload was for its adjusters on a daily basis and to recognize trends, such as how many pending claims were open.

To solve the problem, Fosnaugh indicates, the carrier copied the data from the AS/400 to the SQL server and built staging tables on a monthly, weekly, and detail level so the claims unit could get a snapshot of how many pending claims a certain employee had four weeks prior compared with that day. "Through these new reports, we are able to push open claims from one adjuster to another," he says. "Previously, we didn't have that visibility."

An internal best practice involves helping people understand what the data is for, explains Rod Travers, senior vice president of the Robert E. Nolan Co. People on the data-entry end sometimes don't have a deep understanding of what they are entering. "You are not going to expect pure data-entry people to understand rating tables or coverages to any great degree," he asserts. "But for them to have a more well-rounded understanding of the data they are dealing with is beneficial down the road. They can make contextual judgments about whether what they are entering makes sense or not."

Travers views this as a training issue. "It's not once and done," he says. "It's a continuing process to keep people familiar with what the data is used for and the ramifications of why it is important."

Buluswar's team has developed capabilities to access and understand the different data sources, but they've learned too much gets lost in translation. "There often is a big gap between an IT view of what the data is vs. the business analysis view of what the data needs to be," he says. "The thinking each of those groups brings to the table sometimes is too wide to be able to manage."

As probably the largest consumer of data within Farmers, Buluswar and his team see issues around ensuring consistency. Rather than focus on such issues, though, Buluswar points out, "We have focused more on our team's ability to understand, manipulate, and capture the right elements of information in order for us to be as effective as we need to be." It might be different if his team had similar needs for the data as other business units, but Buluswar explains, "I have found the needs of my team are much deeper and wider than pretty much ever was conceived within the organization."

The next step at Brotherhood is to look at its strategy. The carrier is leaning toward high-level executive reports, which will have the secondary benefit of offering underwriters the lower-level data needed to track business. "We are taking a direct copy of the data from the AS/400 and building a data mart–basically a big staging table–with all the data we are going to need," says Fosnaugh. "When we do the reports, it is easy to jump in and get your data out."

"Data quality goes across any user group or intended purpose," says Travers. But carriers have learned some data is more critical than other data. "Obviously regulatory and financial information tends to get a higher degree of attention," he continues. "If you have someone's address not quite right, that's a different severity of data quality problem. Data quality is the most fundamental of best practices."

To keep that data clean, Travers believes carriers need to minimize the number of ways the data can find its way into systems. "You need to keep the number of input paths as few as possible," he advises, adding it's preferable to have a heavier hand in auditing the data from a Web self-service mechanism, as opposed to an experienced employee entering the data. "The second person is going to have a greater understanding of [data quality] than the person entering it on the Web," he says.

No one can argue against having clean data, but Danielson observes a gap between acknowledging the need and actually doing something about it. "When data is initially entered, you need the quality edits to make sure the data is accurate and related in a powerful way," he says. "People talk about how important quality is, but it's a very expensive [issue] to address. You can't have one person to clean it up. It usually involves one person calling in many others."

Many companies are implementing better data solutions that allow them to manage the distribution channel and help them cross-sell, observes Danielson. "People are spending on more robust tools to understand operational information–performance metrics," he says. "As people focus on cost reduction–more in life than in P&C, given the nature of the cycle we are in–you are seeing people ask for better and more accurate information."

Farmers imports some data from outside the enterprise, but for the most part, Buluswar finds a good chunk of his team's needs are met by internal data. "Fill rates tend to be weaker the more you go outside," he says. "The implementation of any insight or strategy that requires external data pools tends to be more complicated, so we do that only if there is a compelling reason and if it allows us to manage issues around that strategy."

For some issues, Buluswar has found it's not whether data is available but rather whether the data is structured or convenient. Quite often, he finds the data is not structured. "That means the amount of energy we have to spend in order to extract it to a format we need is enormous, but we've made some pretty big investments that are mitigating the scale of that," he says.

Buluswar sees the big issue with unstructured data being how the data is collected in the first place. "In general, I won't say the quality of the data is bad; it's the quality of the person or the people responsible for extracting [the data] that has a lot more variance," he says.

Tools have been developed for aggregating unstructured data and making it searchable or consumable, notes Travers, and that has increased the focus on unstructured data. "The question is how much valuable information is in that unstructured data," he says. Carriers have to set some boundaries on what it is they are looking for to be able to guide the degree of effort needed to categorize or organize the unstructured data. "Otherwise you are doing free-text searches, and you have no way of prioritizing the unstructured data so you can find what you are looking for," says Travers.

Unstructured data creates other problems, such as retention. The mantra Travers hears from companies and legal experts is "store only what you need." Retention policies are being reviewed to make sure companies are storing only what good business practices and the law bind them to store. "Beyond that, don't store it," he says. "Make sure you understand the significance of something you have stored in an unstructured field."

One reason unstructured data is gaining attention is because a corporation's general counsel is being asked about the company's backup and data retention policies. "The definition on how we retain information is being driven by the attorneys," says Danielson. "It's treated in the realm of regulatory compliance."

Many companies have adopted the attitude of doing whatever it takes to satisfy regulatory agencies, contends Danielson. Carriers question the need to spend more money if the regulators are not hounding them about a particular issue. "Ultimately it could come around and bite them, but that's a calculated risk," he says. "Sometimes the fines are less expensive than building the solution. People factor that in."

Many organizations abdicate data management to their systems, according to Travers, leaving the claims system to manage claims data, policy systems to manage policy data, and so on. As more systems are being purchased rather than built internally, he believes little thought is given to how the data is handled, stored, and managed.

The next tier of insurance carriers includes those that manage their data at an integrated level through a data warehouse or an operational data store where they have established relationships between their various pools of data and gone through an architecting process to determine how they will store and use their data.

The third level of insurance carrier takes the data warehouse to the next step. "They say, here is how we are going to take the data and put it to a strategic purpose," says Travers.

For example, these insurers want to understand things such as their trends to a coastal exposure. "They want not only to understand claims data but correlate flood plain data, weather pattern information, underwriting information, and demographics," he says. "That's a very sophisticated thing to be doing, and not a lot of companies are taking it to that level."

Robert E. Nolan Co. senior consultant Eugene Reagan offers his view on the issues carriers need to address in putting together a comprehensive data management strategy.

First, include the following elements:

o A data management function, if scale warrants.

o Information architect.

o Metadata repository.

o User profiling/requirements.

o Data warehouse, operational data store.

Next, define the life cycle of data:

o When it is created.

o Instances when it is accessed/consumed.

o Instances when it is modified.

o When it is discarded or archived.

Finally, implement a data quality program:

o Introduce data quality concepts and procedures.

o Educate employees on the meaning of the data itself and ramifications of poor data quality specific to that data.