For decades, companies behind large, integrated transactional systems such as Enterprise Resource Planning (ERPs), Electronic Health Records (EHRs) and Managed Care Applications have disseminated the notion that having all data in one database brings multiple benefits. At the transactional level, this argument is supported by benefits ranging from infrastructure cost avoidance to higher data integrity and fewer data governance issues. When it comes to analytics platforms, the “one database” model is extremely limiting for organizations and does not produce the benefits seen at the transactional data level.
As someone who has been managing health information technology for over 20 years, it’s important to me to explain why the “one database” argument loses its validity when applied to analytics platforms across multiple industries. What remains is an obstacle to adoption of advanced analytics (descriptive, predictive and prescriptive) resultant from the lack of data in sufficient volumes, with the necessary level of standardization, technology and focused research.
Infrastructure Costs
Transactional data is usually represented in highly normalized relational databases optimized to support online transaction processing. Very often “reporting” databases, Operational Data Stores and Data Warehouses are created to support data and analytics initiatives. PointRight’s analytic initiatives center around using multiple datasets and sources to research and create advanced analytics solutions that further clinical, financial and operational objectives in the post-acute space.
Technology has evolved to a point in which modern advanced analytics platforms can be cloud-hosted in highly cost-effective multi-tenant environments. This allows companies to build affordable, powerful and focused analytics services leveraging multi-tenancy to achieve results not possible in traditional transactional single tenant environments.
Data Governance
Many problems created by multiple “sources of truth” at the transactional level can be avoided by having ERPs, EHRs and other large transactional systems store the bulk of transactional data. Because, by design, analytics platforms ingest controlled copies of transactional data, they are never the “source of truth” for anything other than the derived data elements presented as valuable insights and not present in the source transactional systems.
Analytics vs. Reporting
While well integrated transactional systems are perfectly capable of meeting an organization’s basic operational reporting needs, the availability of standard data in large scale environments with focused research allows for a deeper understanding of underlying datasets. This makes it possible for the development of adjusted measures (risk, population geography, etc.), meaningful benchmarks and accurate prediction of important outcomes. In most cases, this is simply not possible with the “one database” approach.
More specifically, a transactional system gives you reporting, which is really like looking in the rearview mirror. To be able to look through the front windshield, to be able to compete in a value-based environment, this is simply not sufficient. Cloud-based, software as a service (SaaS) analytics platforms, leveraging machine learning and AI quickly and cost effectively add capabilities out of reach for most organizations, even large well funded companies. Issues including lack of data standardization, small data samples, insufficient funding and resources all impede the development of high-quality predictive and prescriptive analytics.
Integrated Transactional Databases and SaaS Analytics Platforms – The best of both worlds
Organizations do benefit from keeping transactional level data in as few databases or repositories as possible. Increasing the level of integration and consistency of transactional data should always be part of the technology strategy. When it comes to analytics, especially advanced analytics, leveraging SaaS platforms in mission critical areas delivers unique value without the downsides associated with multiple transactional data sources. With the spread of AI, the ability to quickly deploy data-driven initiatives, products and services is becoming a condition to “play” in most industries. Having run large enterprise-wide analytics programs in healthcare, and now leading technology at PointRight, this CIO can emphatically say that the ‘One Database’ model has a role and combining SaaS analytics platforms creates unique capabilities that are better together.
Client Login
