Data strategy at an organization follows multiple paths:
> Data Governance and Definitions
> Data Process, Quality, and Augmentation
> Data Warehousing
> Reports, Dashboards, and BI
> Predictive Analytics
> Big Data and External Data Sources
When I talk to insurers about their data strategy, I like to assess how far along they’ve progressed on the above paths. The exact breakdown (or naming) of these data strategy paths can vary from company to company depending on priorities and opinion. But the key is that they all matter and, just as importantly, that they happen in parallel rather than in series.
There are two problems I find when diving into the strategy details. Either (a) some of the critical paths of data strategy have been ignored or, the opposite issue, (b) some of the incomplete paths are being treated like roadblocks to other progress.
The first problem is pretty easy to understand. If an insurer focuses on just data warehousing and reporting (one of the most common scenarios) the data will never really represent a single-source-of-the-truth, the reports and other BI will always be in contention, and there will be lots of greater values left on the table. For another example, if an insurer puts all their effort into predictive modeling, those models will never be as deep or precise as they could be with better data for analysis. It’s not a surprise, though, that this kind of uneven approach happens all the time; a balanced data strategy is difficult and few insurers have the resources or skill in all areas. The different paths require various technological expertise, while still others require political will.
The second problem, on the other hand, requires rethinking how these different data strategy paths interact. Up above I’ve lined them up in what seems like a natural order: first you need to have some kind of governance group that agrees on what the data means, then you need to have a process to clean and flow the data through the different systems, then you aggregate the data into a warehouse, then you report on it, then you analyze it and build predictive models, and only then do you think about bringing in big data to the picture. It makes logical sense. But it’s also wrong.
The reality is that an insurer can work on all of those paths in any order and/or simultaneously. You don’t need a perfect data warehouse before you start thinking about predictive modeling (in fact, there are plenty of third-party vendors who help you skip right to the predictive models by using industry data). You can run reports directly off your claims system even if it’s data in isolation. Nowhere is there more proof of this than the fact that most insurers hardly have any data governance in place but have still moved forward in other aspects of their data strategy. That doesn’t mean a company should ignore the other paths (that leads to the first problem), but it does mean progress can be made in multiple areas at once.
What’s important to understand is that all these different data strategy paths enhance each other. The further an insurer is down all of them, the stronger each one will be, leading to the best decision making and data-driven actions.
So it’s always good to step back and take a look at the data across an organization, assessing each of these paths individually and seeing what can be done to move each forward. A good data strategy has a plan to advance each path, but also recognizes that no path needs to block another depending on current priorities.