Many companies that lack a data science team have Business Intelligence (BI) teams set up to provide reporting and dashboards. These BI functions are frequently where companies turn to in hopes of expanding into other analytic techniques. Unfortunately, the skillset contained in many of these teams does not always match the skills necessary to perform machine learning, artificial intelligence, or other advanced analytics functions.
Let’s talk about Business Intelligence
Let’s talk a bit about BI, or what is commonly referred to in the analytic pyramid (Figure 1) as Descriptive Analytics.
Whether viewing what the raw data looks like with histograms, using visualizations to see trends and patterns in the data to provide insights, or providing transparency in how a model is performing, visualizations are a foundational piece of what data scientists do on nearly every project. That said, the way in which many BI teams work is not aligned with world-class data science functions and is frequently born from following an IT project model. These business intelligence teams are requested to use BI tools to build specific reports and visualizations and are set up as a factory of sorts, churning out data analysis to specification. The reports are then consumed by the requesting organizations, or at times, not consumed.
In general, the measures of success of these ‘classic BI teams’ have been around how quickly they can produce a dashboard and how nice they look.
We could built this out a bit more. Maybe about how they think that the analytics box is checked because they have BI but they’re missing some key components – if you got started on your analytic journey with a BI team you’re on the right path, but you’re not at the end of the road?
Data Science vs. Business Intelligence Teams
The data science team that performs business intelligence has a vastly different goal in mind. The data science team is working to create insight, or even better, to change a business outcome. A visualization created by a data scientist might include an e-mail to alert the business that action needs to be taken, with the necessary graphs, charts, or data analysis for the context to be understood. The project likely wouldn’t have had a requirements document or specification, but instead would have started with a problem. It is quite likely that the business user would have no idea what the outcome would look like when they engaged the data scientist, and infact, if the business user could draw what they wanted it to look like, a data scientist would likely shrug their shoulders and ask, “then why do you need me”? Just like going to a Six Sigma Blackbelt and suggesting that you have a problem and know exactly what to do about it would yield a shoulder shrug and confusion.
Differences Between Business Intelligence Analysts and Data Scientists
So yes, BI is a critical component of data science and is typically a part of nearly every solution, but most BI teams are not actually engaged in the art of data science. Many BI teams are taking orders to build data visualizations versus providing insight and outcomes. To change this, there are typically many underlying factors that would need to change:
- Charter: The team’s mission would transition from delivering BI to helping solve the key problems of the organization.
- Organizational: Many BI teams are part of an IT function that takes orders and bills out the service. The first change would be to move these organizations out of IT and change the model to match world-class data science organizations that are centrally funded with reporting to the CEO, COO, or CFO.
- Skillset: The BI team would need to either develop or bring in those with the full spectrum of skills required of world-class data scientists. From problem formulation and domain knowledge to statistics and modeling techniques to supplement the BI talent they already have.
Are you part of a Business Intelligence team? Do you think you could deliver more impactful results with these changes? What other changes would you like to see?