Written by Sam Tyler (senior consultant in our healthcare sales & marketing intelligence practice):
What is Advanced Analytics? How do you implement it? Does it really add any value? Questions such as these are beginning to plague healthcare business leaders across the nation.
First, some definitions. Business Intelligence has been around for several decades (longer if you include former names such as decision support). It refers to the creation of reports, dashboards, or scorecards that explain what happened and why it happened. It also refers to the acquisition, processing, and storage of data within a standardized format.
Exploratory Data Analysis is the mining of data to look for relationships within the data. Tasks such as segmentation, frequency analysis, and correlation analysis fall under this umbrella. This step often precedes predictive analytics or prescriptive analytics.
Several people confuse advanced analytics with predictive analytics. But, whereas advanced analytics is a general term, predictive analytics specifically focuses on determining an outcome based on historical data. There are several algorithms that accomplish this. The trick is knowing which algorithms fall into which scenarios. For example, I wouldn’t use an ARIMA model (predicts based on time alone) for a prediction on leads value.
Prescriptive analytics tells you what you should do. It can involve heuristics to troubleshoot revenue cycle problems or other operational issues. It can also prescribe emerging markets to pursue. This form of analytics combines the data with the business knowledge to make the best guess as to where to go next.
On top of this is artificial intelligence: the automating and “learning” of the models to self-correct over time. All of these steps beyond business intelligence fall under the umbrella of “Advanced Analytics”. (Although, arguably, artificial intelligence does not as most projects fail to consider how to leverage AI to maintain models. There is also the consideration that AI is the step after the initial project.)
Second, implementation. At a high level, it’s simply (1) Analyze the data for relationships, (2) Create an initial model, (3) Test the model, and (4) Apply the model to a test situation. At a lower level, things begin to break down. Most organizations are not ready because they do not have the data organized or even identified as to what would create the most value. And, even within an industry, there is no “one model fits all” that all the statistical models in the world just latch on to. It’s unique to each company.
Finally, technically minded people latch onto the solution and the model and fail to provide the final step to realize value: the pilot project. You see, usually by the end of all this, there are several possible changes to make or opportunities to pursue. But unfortunately, the project usually stops there. The next step would be to create an experiment to see if the changes identified will hold any value. If it’s successful, they can be rolled out to the whole organization or modified to address unforeseen concerns. Then, and only then, will you see real value from advanced analytics.
In summary, the term “Advanced Analytics” is often overused and has a lot of hype around it. But with the correct skill set, appropriate amount and structure of data, and correct oversight, it will add value.