Insurers must resolve their data quality crisis

Insurance companies must know their customers' profiles, preferences, claims status and histories if they are to personalise their relationships with them and offer new, pertinent products and services aimed at increasing client value. 

A single view of the customer is the ultimate of insurance process and systems programmes. It leads to fluid customer interaction, optimised sales opportunities and increased value that drive profitability, revenue and customer satisfaction.

But 88% of representatives at a recent Actuarial Transformation Roundtable organised by the Insurance and Actuarial Advisory Services practice of Ernst and Young said data management issues made financial data delivery unreliable. 56% were concerned there was no data governance team and 67% that there was no formal data management programme.

Data quality is no different in the insurance industry to any other industry. All businesses require good data quality for producing trusted and accurate information for decision-making.

And the insurance industry faces many of the same challenges as other industries in striving for that, such as legacy systems. Those usually have much dirty data using old or antiquated data rules. That negatively impacts data quality projects because the old data must be understood, mapped and transformed to conform to new rules. Interfacing data quality applications and engines into legacy systems using old technologies and software can also be a challenge.

New data dictionaries must ensure that the mappings and transformations convey the correct meanings of the data to the end-users and legacy system custodians.

In addition, conformance of key data items, reporting hierarchies and business processes across divisions, departments and companies to a common format or convention and practice is an age-old problem. It makes the task of data integration and standardisation, fundamental to arriving at "one version of the truth", very difficult. However, on a more positive note, it brings together the people from these separate sections to discuss and agree on common standards that are healthy for the organisation.

Volume of data

One area insurers are better off than other industries is the volume of data they process. Insurers experience average volumes unlike businesses that log daily transactions, such as banks and telcos. Insurers usually act on monthly or quarterly time periods. Data volumes impact the size and capacity requirements of systems, but they can require a peculiar extract, transform and load (ETL) process. However, the availability and timeousness of large data volumes can require automated data quality tools to be configured and used differently, such as for batch processing.

Taking care of all of these facets means insurers get trusted information delivery for decision-making. Nothing is more important than this benefit. The ICT industry exists for one purpose alone and that is data processing for communication and or publication. Poor quality data can result in a great deal of damage.

So where do insurers start? The Association for Cooperative Operations Research and Development (ACORD) standards are a good jumping-off point, particularly for companies without any standards that are looking to implement those that are up-to-date and accepted industry-wide. ACORD offers a comprehensive set of general and specific data standards that are aligned with the insurance industry. They also offer tried and tested solution templates for data and consider the latest regulatory requirements. Companies can also use ACORD to gauge their own standards and practices.

It is common knowledge that standards help to manage and enhance not just data, but processes too. Conforming to standards leads to easier data and information integration and sharing. Data standards are guidelines or "place-holders" to achieve commonality. But to implement standards is not a straightforward "buy and adhere" exercise. From the outset, standards are born of or evolve from companies business practices. Companies must first agree their business strategies based on industry best practices and regulatory requirements. All other strategies are guided from business strategies; for example, the collection of data, information and technology strategies. These sub-strategies prescribe the exact requirements for standards. Successfully outlining these relies on skilled and experienced employees, such as business and systems analysts and data and systems architects.

Deploying standards from an organisation such as ACORD leaves little for insurers to do. There will be a certain amount of customisation depending on the business. Companies attempting to develop and implement their own standards, however, could spend as much investigating and creating the standards as it would cost to buy them in an off-the-shelf product.

But there are a limited number of packaged data quality tools available to help them achieve that and some of them offer limited functionality while others offer more and conform to industry-wide practices and standards. Those developing their own data quality rules and standards, dependent on budget, data volume and the state of the data, will use programming software tools and development languages that they may already have, such as standard SQL, Visual Basic, C++ and others.

The entire exercise becomes an examination of time and cost of development versus cost to buy and time to adapt.

Whatever the choice, every data quality project is going to hinge on metadata. Its the definitions of data and processes. It gives a data item context, such as the type, meaning, location, usage, and its relationships. Without metadata, data stands isolated and cannot be properly used or understood. Standards are only achieved by using and managing metadata. Its the "place-holder" for data and process standards that can be referenced, communicated, controlled, managed and understood from a central point of view and strongly supports the requirement for data conformance and commonality.

Data rules and standards also cannot be implemented without a champion. However, creating a formalised data management team is the best way to achieve that, as it will be driven from a business process perspective, eliminating the dependence on individual employees.

The quick guide to getting a data quality project off the ground is:

* Establish a formal data management function consisting of data architects, business and systems analysts who understand the business strategy the function must be fully sponsored and sanctioned by top management;

* The data management team must identify, based on computer applications and systems, all the data items necessary to operate or turn the business. They must also identify manual and external data needed for operational and management purposes, for example: exchange rates;

* They need to assess the business processes and procedures to confirm data usage;

* They must determine and agree on a standard data naming and storage convention between the different systems, including and mappings and rules;

* The data management team must agree all necessary business rules for the data and agree default values for missing data;

* They should formulate data transformation and correction rules for all key data items and non-free- text data items;

* Once completed they must document and publicise the data quality standards and rules; and

* Finally, they need to apply these rules and standards into the operational applications and business intelligence systems.

But data quality doesn't end with a successful project. Its an ongoing process sustained as an operations process as part of the data management function.

By Mervyn Mooi, director of Knowledge Integration Dynamics (KID)