Look under the hood of a car and complexity stares back at you. There are hoses, wires, clamps, fittings, bolts, components, several different liquids and gases, containers, pulleys, belts, fans, lights and fuses, and the real gem — computer technology. And, of course, there is an engine built from hundreds of parts that you can’t see.
An early 1900’s look under the hood was something entirely different. A novice could understand most of it. But, then again, the early car broke down frequently and didn’t generate much horsepower. Today’s complexity improves performance.
As insurers, we’ve grown our data use in much the same way. We now have the potential to gain much better information and value from our data (more horsepower), but we’ve exchanged data’s simplicity for hoses, belts, wires and engines that require ground-up understanding. Similar to cars, we need the computer intelligence to help run it efficiently.
This is what we have been doing for the past several weeks — exploring the basics that will help us construct the best data platforms — giving us high performance and helping us to manage the complexity. In last week’s blog, Get the Data Basics Right: Part 2, Data Infrastructure, we looked at how a well-constructed EDW, built on an effective underlying data model, gives insurers better Business Intelligence and a dynamic framework for integrating supplemental data sources. You can find the full description in our recent thought-leadership report, Digital Insurance 2.0: Building Your Future on a Robust Data Foundation. This week, we are stepping out of the warehouse and into the front office. How can we construct our analytics infrastructure in ways so that we will receive valid, timely answers to critical business questions?
Key components to operational analytics
There are five key components required to successfully perform operational analytics at the enterprise level:
- Trusted and secure data
- Insurance subject dimensional data model
- Integrated data
- Business-driven self-service UI
- Social collaboration
Trusted and secure data
There are many excellent visual display solutions that exist today that have been very successful for carriers at the department level, but their Achilles heel is the ease with which they can bring data into the organization. Carriers need to be able to trust the data underlying the reports they are examining. This means that the data must have gone through data governance for enterprise data quality checks, business rules validation and approved semantic definitions. Additionally, it requires a barrier to access. No one should add to nor modify the data, or else one changes the meaning of it. Managers need to know if the data is strictly from the data warehouse which has been verified against company standards or if the data has been added to or modified. This can only be accomplished by a solution that does not allow manual modifications to the data or definitions and has a very strong manual data governance process on reports.
The security placed around the data is also essential for creating trust in what it is telling the user and which users can see which data. An analytics solution must be able to provide security along several fronts: access to reports; create/modify privileges for dashboards and reports; and the data itself. Some reports may need to be restricted to certain groups or even individuals. Insurers today have literally hundreds of reports that are generated, but never used. Insurers must ensure that reports and dashboards conform to a certain enterprise standard; that security is appropriately defined for them; and the creation of a plethora of redundant or nearly redundant reports is prevented.
Insurance subject dimensional data model
A third normal form (3NF) model is excellent for an enterprise data warehouse and can be used for operational or listing reports. These reports are generally predefined, static in nature and allow business users to filter the data. To efficiently enable operational analytics (where business users can drill down, drill across and drill anywhere to discover root cause analysis) an insurance dimensional data model must be used. Many data visualization and BI tools can either sit on top of a dimensional model or provide the ability to create a dimensional model between the data source and the user interface. Dimensional models allow for accumulating snapshot tables for efficient summarization, as well as facts transaction tables to enable users to get to the specific policy(ies) or claim(s) that can cause an anomaly.
Not only do carriers require root cause analysis of their policy or claims data, but they also want it to be integrated, to enable business performance reports such as loss ratio analysis. It is not enough to only provide policy or claims operational analytics. To track and explore business performance, the policy and claims data, along with the associated party data must also be fully connected.
Business-driven self-service UI
To efficiently allow business users to get the insight they need at the time they can use it, there must be a business-user-focused, self-service BI front end. Many carriers today either have to rely on a special reporting or data science group to do their data analysis and reporting, hire technical experts on the selected reporting or data tool chosen, or provide their reporting requirements to an IT or reporting team that 4-8 weeks later will give them the desired report. In many instances, this leads to frustration in the time delay, but also having to repeat this process for follow up questions. Business users would like easy access to the data and reports that they need in order to follow their train of thought without having to depend on others. This ability to have the right data or analysis at the right time empowers business users to truly improve the business.
Finally, business users in many instances wish to collaborate when researching a data or report anomaly or searching for business insight. They may wish to converse with peers or assign tasks for others to follow up on. Business users would like to be able to have a history of these interactions and assignments connected to the report, as well as determine how private or public these conversations should be. Long e-mail trails with numerous out-of-sync replies, forwards and cc’s seem to be common, but suffer as a solution. Not everyone along the analysis journey may have been involved from the beginning, and new resources may have been brought on or replaced. Final documentation or annotation should be directly tied to the report so any stakeholder who wishes to understand the anomaly now, or even months later, or if someone desires to remember the root cause, can find it logged within the report and not lost in a forgotten document.
Majesco Business Analytics (MBA) incorporates all five of these enterprise analytic requirements. The data flow or ETL (Extract-Transform-Load) process is completely automated and data descriptions are provided for both IT and business users to fully understand the meaning of each data entity and calculation. Only selected administrators are able to make any changes to the data, metadata or data flow, so business users can trust the underlying data in the reports that they are viewing. MBA uses a robust and proven data model that is based on a Kimball Insurance dimensional model to enable root cause analysis. The MBA data model and implementation integrate policy and claims data across shared attributes, such as producer, LOB, peril, coverage, and more, to enable drill-anywhere capabilities.
The self-service BI UI also supports social collaboration, allowing users to share conversations, either publicly or privately, assign tasks, and annotate reports for single dates or date ranges, again, either publicly or privately. The self-service front-end provides an optional workflow to support carriers that wish to have reports reviewed and governed prior to being available for public consumption. The front end also allows for data extraction to various document formats, based on security privileges set up by the carrier.
Analytics Infrastructure is a competitive differentiator
Digging into the details (looking under the hood), analytics can seem internally-focused. It is crucial for analytics users to keep their eyes on the end results — an organization focused on competing for customers and markets. Competitive dominance is no longer achieved by operational efficiency, lower prices, massive advertising, large internal systems, or channel loyalty. It is achieved by anticipating trends and pivoting quickly to create and capture the economic and competitive opportunity through a new business model built on data. Insurers that succeed in capturing the opportunity will be those that show strategic courage and forward thinking to redefine their business models, processes, products, services and channels by leveraging data and analytics in innovative ways, becoming data-driven and laying the foundation for Digital Insurance 2.0.