Embedded Analytics
I have spent most of my career focused on data analytics, founding a data warehouse company in 2000 and later as CEO of a predictive analytics company.
A comment I recently heard on a webinar about data quality was “no company designs data for analytics, it is designed for operations”. This is typically true, but we have found that a feedback loop from analytics is very desirable for most systems, and necessary for digital insurance.
Back in June 2020, I wrote about Analytics in RiskDX. The key takeaway from this was that we have embedded the analytics using Embedded PowerBI giving users the flexibility to slice and dice data without the need for additional software.
Since then, we have continued to build our analytic suite and iterate the data design used for operations. This process of continuous improvement is important and only possible within an integrated operational and analytic environment. Three stages of analytics development can be summarized as follows:
Deployment of automated analytics, regularly update analytic tables and dashboards or reports. This should also include de-personalization of data so that the analytical data can be retained longer than you need to retain operational data.Structure operational data to include analytical data. The main purpose is to embed context, for example the question that causes an application to be referred is not important as it comes at the end of a string of other questions, what you need for analytics is the reason for referral structured in a way that lessons can be learnt.
Structure operational processes to integrate analytics. A major reason for analytics is to improve business processes, so there should be a way to incorporate analytics into operations. By building this before the analytics is carried out, the algorithm can be substituted with A/B testing scenarios to directly feed analytics.
These stages also place different demands on the analytical environment. Automated analytics needs an environment that is intuitive to users and lacks barriers to entry such as additional license fees. As companies get deeper into analytics, availability and understanding of data becomes more important as users may use more specialized software specific to the task. This leads to the need for data extracts and good quality metadata.
RiskDX supports a full range of analytics scenarios and continues to evolve. We recognize that supporting analytics is as important as the operational functionality.