High Quality Analytics from High Quality Data: Data Quality Review

December 13, 2016 / 3 min read

High-quality analytics are critical to evaluating and managing risk. But you can’t produce high-quality analytics without high-quality data.

That’s why at ISO, we work hard to help ensure that the more than 2.8 billion detailed records of premiums and losses we receive from insurers every year are of the highest quality before the information is analyzed by regulators and customers.

7355

A primary need is to have data that’s an accurate representation of reality. As a result, during our data quality review, we focus on two key indicators: distributional shifts and balance.

Distributional shifts: Large changes in the premium volume reported for certain fields, such as class or subline, may indicate a reporting error, possibly resulting from system changes. Similarly, large shifts in losses from one type of loss to another (for example, from bodily injury to property damage) may also indicate an error. There are some variables where shifts may be expected to reflect a change in the market or a specific book of business, but for many others, we would not expect any significant volume changes.

Balance: We also review the balance of premium and losses at an aggregate level. In some lines of business, losses are more volatile, and a large loss ratio may not be a result of misreporting. It may be the result of natural disasters or simply adverse development. But we look carefully at each case to determine what’s behind the change in balance.

Four Common Errors in Insurance Data Reporting

During the data quality review, we find four issues come up the most: premium and loss mismatches, large volume changes, exposure misreporting, and endorsement errors.

Premium and loss mismatches: Premium and loss mismatches are frequently seen when loss coding doesn’t match the premium coding of an underlying policy. For example, a territory or ZIP code may be reported incorrectly on losses and correctly on exposures, resulting in a mismatch that could skew territorial experience and loss cost–level changes.
Large volume changes: Large volume changes in fields such as protection codes or territory codes are common issues across various lines of business. In some cases, these issues may be caused by programming changes while implementing a statistical plan update or even a misinterpretation of the update.
Exposure misreporting: Exposure misreporting is a common issue seen across multiple lines of business. Depending on the line of business, some classes require exposures to be reported in hundreds or thousands of dollars, rather than reported as units. Every quarter, we catch a handful of exposures being misreported for specific classes. Since exposures play an important role in developing loss costs, it’s crucial to report this field correctly.
Endorsement errors: Reporting of endorsements can also lead to extensive misreporting. When extra endorsement coverage is provided for a single policy, the ISO statistical plan generally requires two records to be reported: one with the bulk of the premium for the underlying base policy and one with additional endorsement premium tagged with an endorsement code. A common error seen in our reviews is the entire premium for a policy reported under one record with the endorsement code.

Data Inquiries and My Data Quality

When we encounter data quality issues at ISO, we send an inquiry to the individual insurer, explaining what we’ve found. That often leads to complex and lengthy discussions between the insurer and ISO as we work to resolve the issues.

To streamline the process, we’ve introduced a new self-service portal in ISO’s Statistical Web Services called My Data Quality. The portal allows you to view and manage your data quality cases, performance evaluations, and annual verification requirements. You can view your inquiries, including problem descriptions and supporting exhibits. The portal also includes a comment feature for each case so you can easily send us updates or questions, allowing us to resolve quality issues more efficiently.

If you have any questions about My Data Quality or about how we can help you report, benchmark, and analyze your data, please contact Dennis Huang at Dennis.Huang@verisk.com (201-469-2184) or Richard Morales at Richard.Morales@verisk.com (201-469-2272).

Dennis Huang and Richard Morales work for ISO. Dennis is a Manager of Analytical Data Management and Richard is a Manager of Statistical Data Management.

High Quality Analytics from High Quality Data: The ISO Data Quality Review Process

Four Common Errors in Insurance Data Reporting

Data Inquiries and My Data Quality