What's So Funny About A Single Version of the Truth?

“Single version of the truth” has always been the cardinal virtue for insurance data analytics. There was a time when we lacked data, but now we’re awash in it and struggle to know which data to trust.

We hear this theme a lot.  There are three common storylines:  (1) too many reports with substantially the same metrics and dimensions but arranged differently because they were created for different users at different times by different authors, (2) duplicative departmental data hubs (Claims and Billing, for example) that integrate the same data using different business rules, and (3) dashboards that show different values for the same metrics though few know why.  And, of course, the storylines get co-mingled.

P&C insurance is a data business, yet we’re veracity-challenged. The veracity challenge hangs over two enterprise data management imperatives, which are as conflicting as they are unyielding: governance and control on the one hand versus self-service and agility on the other.

What’s an analytics leader to do?  Here are some promising enterprise-class solutions that we’re seeing.  None is a panacea, but each nudges your company toward a single version of the truth:

Report Rationalization and Modernization – consolidating and eliminating reports to radically reduce their number produces powerful benefits.  Less is more.  If you have not done this recently, 10X reductions are not uncommon.  Some top tier P&Cs are pushing the envelope to get different LOBs using the same report and management metrics – claims productivity, for example -- where formerly they each had their own.  And, a modernized report or dashboard will deliver a more information-rich and analytics-driven user experience.  Rationalizing and modernizing reports elevates analytics and reduces technical debt at the same time.

Governed Data Products combine two future-facing data management concepts:  (1) Data products are autonomous subject area-specific data stores.  There is no limit to the number of decoupled data products or data product producers. (2)  Governed means that the data product’s integration rules, metric definitions, dimension conformance, etc. support the business at large.  Governed data products are not silo-specific; rather, they’re enterprise assets.

Data Marketplaces allow corporate data consumers to find, understand, and use internal data products.  The data marketplace is an online store (sans remuneration in most cases) that allows data consumers to meet producers in a self-service model.  AWS and Snowflake, for example, have marketplaces for external datasets. Internal data marketplaces encourage re-use and continuous improvement because they include a feedback loop between consumers and producers.    

Data Catalogs are not new, but adoption is accelerating. While data sets and reports multiply, the catalog can be the “single pane of glass” through which a common understanding is achieved and knowledge is shared. The data catalog connects to various source types, including relational databases, file systems, and analytic and reporting tools. It extracts valuable metadata from each source, from data popularity and usage patterns to samples, profiling, and lineage.

If “single version of the truth” is data management’s cardinal virtue, succumbing to the lure of silo-specific data solutions is its original sin.  We like the enterprise-class solutions above because they embrace the imperatives for self-service and agility as well as governance and control.

By Mike Lamble

Partner and Founder, PremiumIQ

mlamble@premiumIQ.com