Compare your data project to cleaning out the attic. It can be a dirty and dusty endeavor, but once it's done and everything is clean and sorted you discover some of the items that were buried in the corner are not only useful, but valuable possessions.

Unfortunately for the insurance industry, taking on a data project can take quite a bit longer than cleaning out the attic or the basement. Too often, it can be overwhelming. Insurers, instead, just listen to the complaints about the data.

"All companies complain about their data," said Deb Smallwood, founder of the research and advisory firm SMA, Strategy Meets Action. "But when you look at the list of projects they have planned, very few of them are data projects."

Qasim Hussain, senior architect for the technology consulting group X by 2, agrees that everybody blames data. "Data is never good enough," he says. "IT can go in and fix the data, but often the business is still unsatisfied with the quality of the data."

Hussain also believes that once the data is of a certain age, fixing it is almost impossible.

"It was captured at a point in time by a system, and there is no way to fix the data without interacting with an external entity to get the data cleaned," he says.

For example, Hussain works with a health insurer, and when they went back six years into the data they found many names were incorrect or misspelled.

"There's no way to go back and figure if this is the right person unless you call that person," he says. "You try to fix it if certain attributes match. You can try to use some intelligence to clean up the data, but [the carrier] has to accept the fact you are not going to get the data 100 percent correct."

That means carriers need to decide upfront the level of quality they can call "good enough" to move forward, according to Hussain.

"You can't reach perfection, but can business and IT negotiate, given the state of systems, whether 90 percent is good enough? 95 percent?" he asks. "Can we move forward that way? There will be some margin of error in the analytical platform. Historical data is what it is. You can try some resolution, but there is no way it is going to be perfect."

Exceptions to the Rule

The insurance clients Hussain works with for X by 2 are struggling to get data in a way they can do better analytics and gain some business intelligence off the data.

"The concept of data mastery is you really need to come up with a consolidated way of thinking about the data across the enterprise," he says.

Desjardins General Insurance Group is an exception to the rule. The company has been revamping its business intelligence solution for the last few years. Sonia Sevo, vice president of the business intelligence competency center at Desjardins, explains the carrier always looks to its data before making business decisions.

Desjardins did a diagnostic check on its systems in 2008, which reinforced the view that the insurer's systems were built on aging technology and not very scalable.

"It was difficult to meet the current business demands and requirements as they were evolving," Sevo says. "Our need for information on our customers evolves very quickly if you want to keep up with the competition. What we had wasn't allowing us to meet those requirements and establish a competitive advantage within our industry."

Such projects are not as common as they should be, though.

"The majority of the industry hasn't grasped the concept that data is a business process," said Smallwood. "It's a constant problem. You can't clean it and think it will always remain that way."

Think Big, Act Small

Sevo claims the competition is fierce among direct insurers in the Quebec, Canada market where Desjardins operates. The carrier needed to take a bold step.

"We felt rebuilding what we currently had wasn't going to work," she says.

Desjardins, also a direct carrier, instead decided to move forward.

"We wanted to learn from all the different trials and errors that have been happening in the last 10 years in [business intelligence]," says Sevo. "It can be a complete success or a complete failure, and it's a very fine line between the two."

Desjardins chose the Insurance Insight solution from Oracle, partly because much of the design work that's required when building a data warehouse or a BI environment was already completed.

"You have to have the 'think big, do small' approach," she says. "That's easier said than done when you are establishing a new environment with data as complex as it is in the insurance industry."

With the vertical solution from Oracle, Sevo believes the "think big" approach was already completed. "There is some customization because our requirements are particular, but [Oracle] allowed us to cut up our projects and deliver important information and data to our business users in a timely fashion and improve our time to market," she says.

The approach allowed Desjardins to begin implementing its data governance process, ensuring data quality and consistency across the organization.

"It comes with definitions that are standard within the insurance industry," says Sevo of the new system. "It allows us to move forward quickly because we're not starting from a blank page. It's a model based on projects that have been done across the industry, giving us a good idea on what exists already. The comparisons have already been done."

Data Questions

Smallwood believes data cleansing projects involve five steps:

o Accuracy–Is all the data being collected free from errors?

o Complete–Is all the required data being collected?

o Consistency–Is the data captured and stored in a standard format?

o Relevance–Is all the right data being captured?

o Timeliness–Is the data available in time to provide maximum value?

"The best insurance companies understand that strategy is very important," says Smallwood. "At the same time, they recognize, in the end, company results are determined by the sum of all the individual transactions, interactions and decisions they make throughout each day and over the course of months and years. Embedded in their IT systems and in the minds of their employees is a simple, yet extremely important data philosophy: Do it right with accurate and rich information."

Smallwood believes the poor quality of data residing within an insurance carrier's systems is a major barrier to progress for the carrier.

Data that is erroneous, incomplete, inconsistent, out of date, and disparate, she explains, creates a peril for insurers and makes it difficult to gain insight into the problems faced by the business side. Even worse, poor-quality data may result in lost business, increased costs, and even fines for lack of compliance.

Size Matters

Where Hussain has seen success in data projects is with smaller insurance carriers where the complexity of the data was not as large.

"The reward these carriers receive is they are able to understand more about how to improve operations and change business practices to meet the needs of their constituents or to target markets they thought they were targeting effectively, but where data shows there's a low conversion rate for policies they want to sell," he says.

Small carriers don't always take advantage of the tools available, either.

"It takes time, skill and automation," says Smallwood. "There are a lot of tools available to clean data," she adds, noting that "small companies tend to do it manually, though."

Smallwood understands not all insurers are created equal. The top tier of insurers likely has fulltime data teams. "There is always activity to clean and prepare the data," she says.

Typically, larger insurers may have several departments come up with their own definition of what constitutes a certain policy, a certain number, and what constitutes a certain contract, explains Hussain.

"Getting all of them to agree on one definition and to build that definition to construct some analytics on it is a difficult task," he concedes.

Unfortunately, Hussain hasn't seen as many successes with larger insurers. He notes that two clients currently are in the process of undertaking data projects.

"They are trying to get some small targeted structure built so they can answer some immediate questions, and that in itself can lead to some small successes, so when you build the structure for the long term you have people who have seen the value," he says.

"At the highest level of leadership there seems to be a feeling that it would be great to have [better data] and they are pushing for it, but when you get to the actual people who are going to use [the data], I've seen resistance," he warns.

Whatever the approach, Smallwood is certain of one thing: "You have to keep chipping away at it."

Legacy Issues

Another area that hurts many insurers is their systems typically aren't designed to master the data in a way to use it for analytical purposes.

Hussain recounts visits with clients that are using systems built in the 1970s and 1980s and have yet to migrate to the latest policy administration, membership management or customer relationship management systems.

"There is a lot of legacy out there, and those systems were built with a certain design and certain constraints in mind," says Hussain. "You have to build those concepts in before you can tie everything together and come up with a consolidated view. In legacy systems, data is generally not in a convenient form to use."

Project managers have to be aware there are no quick-hit successes that you can achieve early from a data project.

"You need large volumes of data in a consolidated way before you can start doing any analytics on the systems," says Hussain. "You can't build that in two or three months. The enterprise commitment to follow through [on the data project] needs to be there."

Taking Responsibility

Hussain finds the business side usually leads the way on data projects, but he believes insurers need a special mix of business people with an IT mindset, because those who want to do analytics and need a lot of intelligence about their business are not typical business people.

"You need people who know what types of questions will be asked of the system and the best way and form to ask them," he says.

"Questions are hard to articulate in a requirements document, so you need someone who can understand the technology and say: 'If you build this type of a platform it will be to answer these types of questions.' You need a good business partner and a strong IT partner coming together and understanding how to grow the platform so it will serve the analytical needs of the business."

The Full Power

Smallwood maintains proven tools and processes are available for insurers to improve the quality of their data, enhance its value, and turn data into a carrier's most valuable asset.

"Unleashing the full power of insurance data should be on the short list for every insurer," says Smallwood. "It is doable today. The tools are available and not difficult to use. The path to high quality and rich data is not hard to define, and the goal is within reach for most insurers. The investment is one that brings quick reward and keeps on giving." TD