I’m a big fan of home remodeling shows.  They make it look so simple and easy.  Recently I took on a home remodeling project; it turns out, those shows lie!  Maybe lie is a strong word, but they certainly gloss over the truth. I realized that there are a lot of similarities between remodeling a kitchen and transforming a 1990’s data warehouse to an updated modern data platform.

Two years ago, when I started this job as manager of enterprise analytics at Children’s Minnesota, one of the largest pediatric health systems in the U.S., I was tasked with improving  “self service”.  The organization was struggling with increasing demand for data but only a few people were providing reports. However, there was so much work to do on business intelligence (BI) basics that self service felt like the wrong goal. 

Focus on the DW First

After conducting an assessment, it became quite obvious that the first thing we needed to do was re-architect our data warehouse.  This wasn’t something we considered lightly, I’ve been through these projects before and they can be painful at best and career ending at worst.  So believe me, we thought about this a lot before taking on the challenge.  But the reality was that we had a data warehouse that mimicked a transactional electronic health record (EHR).  In 1999 that was sufficient because that was all we really needed, but as time went on and we added other EHR’s and data systems, the data warehouse became a big obstacle to getting data to users. 

Get Outside Help. Our first step was to hire a technical advisor.  I am an analyst by training and while I’ve been doing this work for awhile, I am still not a technologist.  We needed someone that had a lot of depth in the technology platforms and the architecture of the modern data platform (MDP). 

Stop the Bleeding. Our next step was to stop the bleeding.  We had batch jobs that were running out of the batch window (sound familiar?). That was a relatively easy fix, but still required a short proof of concept that helped us vet our technology approach. 

Select a Platform. Then, after months of research, we chose Oracle 12c with in-memory and partitioning to support our need for high user concurrency.  We bought new hardware and implemented the software.  It took about four months, but once that was complete, we had a stable platform that would allow us to focus on our next step; re-architecting the data warehouse. 

Re-architect. Now, we have to keep our legacy data warehouse up and running while we build a new data warehouse that is smarter and faster. As I write this, we are in the first few months of the re-architecture.  As with most healthcare based data warehouses, we are starting with the patient first and going from there. We have taken a very value-based approach to architecture, ensuring that the subject areas we build are tied tightly to business value and the overall strategic goals of the organization.   

Agile Processes. We are also building processes that allow us to improve our turn-around time; we call that “agile data acquisition”, essentially the ability to quickly add new data sources. 

Organizational Alignment. To cement our architecture, we decided to merge our data warehouse and enterprise analytics teams to ensure that there is a seamless connection between what users need and want and what we build. 

Summary 

So we’ve made a lot of progress quickly. We’ve made big technology investments, adapted agile development concepts, and created an enterprise team that will focus on data governance, starting small and building the capability. Everything is in place for us to transform our information environment to create a data-driven healthcare organization. 

The only thing I know for sure about this transformation, as with my home remodeling project,  that despite our best efforts and plans—something will go wrong.  Over the next few months, I’ll share with you our journey, the lessons learned and our successes.  Now, I have to get back to my remodeling projects!