(Bruce Brodie, Josh Schwartz, Michael Kontra, Matthew Krzywicki, Anup Madampath, Sundeep Thakkar, and Scott Busse contributed to this article.)
Fifty years ago, data managers started to migrate data from punch cards to tapes and magnetic discs. The leap in storage and flexibility seemed profound. Yet, there were limits. Once a new system was up and running, it was far easier to do something creative with an existing field than add a new one. For example, the suffix field could be a perfectly serviceable Do-Not-Contact indicator. Companies eventually cracked down on rogue usage, but that field was DNC from the early 1970s through late 1980s (and typically still is today despite several more decades of data).
What happened to the enterprise data cube that was all the rage in 1999? Why are the actuaries the only ones with a functional data warehouse today?
Analyzing and profiling insurance carrier data is like an archeological dig.
From the long paved over ENIAC cow-paths of the 1950s comes the opportunity for new understanding, new insights, real value. Data science and a broad appetite for analytics has been out of the shadows for a while now, and carriers now have to find ways to democratize their data.
But there’s a big problem. Data remains virtually unusable in layer upon layer of systems, sprawling and disconnected. There is no enterprise data model. Management information is cobbled together from one-off efforts of pulling extracts, matching, cleansing, and investigation. There’s little reusability and less trust in data artifacts. Employees champion small wins, but none stand the test of time. Similarly, analytics models built on top of this sprawl are equally distrusted, suspected of flaws, and deemed illegitimate for business decisioning.
And there’s still another conundrum when carriers set about building the data assets they want. When a company finally gets serious about establishing a common data model, central data repository, and data governance system, something else takes priority and a new “waxy-layer” is added to an already too complex environment. We have ample evidence of this from our own archeological studies.
We all know that customers are important. They’re at the center of every strategy. When we study them, we study them as data. (For more on customer service models, please see PwC’s recent
- The company doesn’t understand the data challenge and therefore doesn’t properly fund it.
- The company doesn’t understand the true cost of data mismanagement, ad hoc efforts, and the inability to model and measure the business using data. (Odds are its competitors don’t either, which limits the pressure to get things right.)
- If there’s no regulatory impetus, then it’s viewed as optional.
As the old adage goes, "the best time to plant a tree is 20 years ago, and the second-best time is now." The same is true of insurance data strategy. Carriers need to live up to their oft quoted aspirations to “be more like Amazon.” Data is at the core of what Amazon has accomplished with their technology stack and has been monetized through AWS S3. Data continues to yield a competitive advantage to Amazon as analytics, machine learning, and artificial intelligence continue to grow in importance and adoption. What are other companies waiting for? These are the essential steps to providing a modern, data spinal column for your company:
1. Model thyself. Establishing a Common Enterprise Data Model is the non-negotiable starting point because without it your company will lack a common language to describe itself. Once a data model is established, it can evolve over time, but it acts as a bulwark against ambiguity and data proliferation chaos. The data model should contain a collection of business domains that accurately describe how you do business and measure effectiveness. We’ve learned to avoid generic one-size-fits-all models with an unwieldy number of extraneous domains and fields that have nothing to do with the business they are modeling and create a hugely taxing payload for integration purposes. Instead, it’s preferable to begin with a solid core data model that’s been battle tested and deployed by multiple companies – one that’s suitable to your insurance focus, whether life/health, personal lines, or commercial – and then customized to your unique needs through the translation of business use cases from real life.
2. Jump in the lake. Cloud data storage is now so ubiquitous, cost-effective, and secure that having a Cloud-based data lake to act as the initial Landing Zone for the data coming from your multiple systems of record is a no-brainer. Once your selected source system data -- from all your underwriting, policy administration, claims, human resource management systems, etc. -- is swimming in the lake, your next choice is how you want to logically make the data available to hungry consumers.
3. Window to your world. The Central Data Repository is where you serve your data customers. Depending on the core skills within your company, you might choose a non-SQL option or a relational database that does support traditional SQL queries, or preferably both to support different classes of end user. Cloud-based options again offer the greatest set of capabilities for the least amount of worry and internal shepherding. Similarly, visualization tools best make the data come alive for the greatest number of users.
4. Tend the garden. Once you have rallied your supporters, taken the plunge and brought a new enterprise data store to life, you need to provide a framework for governing it or it will succumb to the same natural causes that plagued every other legacy data store. But you have better tools now to fight the weeds! There’s an enterprise level data model used in every data-related conversation. There’s a populated data store that’s recognized for delivering trusted data to many consumers. Now you need a governance model that enables the enterprise model to evolve to meet the demands of the business. By being responsive and delivering what people need, you will provide a tremendous incentive to play by the rules and not create alternative rogue data stores.
In sum, the steps to achieving a trusted enterprise data store are not overly complex, but the leap from thinking within a complex, ad hoc, ungoverned world to one of order and trust is hard for many carriers (and non-carriers alike) to fathom. Ultimately, companies have to decide that they can’t endure the pain of being an invertebrate anymore and commit to having a backbone – an enterprise data backbone.
