Note: This is part 4 of a 4-part series. For an overview of the series, check out Part 1. For more on how to prepare you data for analysis, read Part 2. If you want to get stoked about Machine Learning, Part 3 is for you.
As a Data Scientist I have trained countless machine learning models, but the most interesting by far is the first model I ever made. I was working for a Major League Baseball team at the time, and our players and coaches wanted help understanding which pitches they should throw to which batters. So I brilliantly answered their question using an ensemble of gradient boosted trees and generalized additive models that predicted the likelihood of various terminal outcomes based on a robust feature space of continuous pitch measures and binary player identity variables.
If you didn’t quite follow that last sentence, don’t worry. The pitchers and coaches who needed the information to do their jobs didn’t either. And when I tried to explain it to them on those terms guess who looked like the biggest idiot in the room…
It was me.
The reality is that Data Science for the sake of Data Science is pretty worthless. You can come up with incredible insight, but if you want that insight to result in meaningful change, it needs to be in the hands of those who have the power to enact meaningful change. Most key decision makers don’t have a background in Data Science, and those who do are probably too busy to try and understand your analysis in-depth anyways. What these end-users really need is something clear and easy to understand that still accurately conveys the insight you have.
This is where Business Intelligence comes in. Business Intelligence just means presenting your data to those who need it through data visualization and reporting.
Data visualization is exactly what it sounds like: creating some type of helpful image based on your data, usually as a chart or graph. Reporting is also what it sounds like: presenting the information regularly in some type of written or visualized report. When properly crafted, these tools will present to your decision makers all of the information they need in a format they can easily digest.
To be clear we are not talking about dumbing things down, we are talking about providing translations. Your end user may not speak Data Science, but most of the time they are in the position they are in for a reason. So if you can equip them with insight in a language they speak, you will see results.
While business intelligence can come in a lot of shapes and sizes, the most popular tool by far is an interactive dashboard. Dashboards allow decision makers to select their own inputs to get the exact chart or information they need at that moment. While dashboards can be coded from scratch, popular tools like Tableau, Microsoft Power BI, Google Data Studio, Shiny, and Plotly Dash exist that allow users to create feature-rich dashboards very quickly.
For an example of what this looks like in an actual organization, we can return to my baseball example from earlier. I realized that if I didn’t find a way to convey my findings to the players and coaches in terms they could understand, my fancy machine learning model would go to waste. Fortunately, baseball players are used to looking at charts, and are given a report before every game that presents all of the information they need in order to be prepared for that night’s matchup.
So I diligently built charts that summarized my model results in an image that players could understand. I then worked carefully with colleagues to make sure these charts would be included in the daily report and that the charts I selected were the most relevant ones available. To complement these charts, I even wrote brief text summaries of key takeaways for each player. Finally, I built an interactive dashboard that allowed users to build their own custom charts based on any combination of inputs. When it was all said and done, I had easily spent just as much time on helping people understand the results of the model as I had on building the model in the first place.
Only after I put in this time and effort did the players begin to change their in-game strategy for the better. In a similar manner, your end users aren’t going to make the right decisions until they have all of the information they need presented in as many ways as they need it presented.
So if you’re looking to get started with Data Science, you cannot neglect Business Intelligence. In fact, the tools mentioned above can be a great way to get started with Data Science. That way, as you get more data and build more models, you will have executives and decision makers that are already used to using Business Intelligence to make decisions. Throw in a little bit of Data Science expertise (like they kind provided by Melgren Analytics), and all of the puzzle pieces will be in place. You’ll have great data, great insight, and most importantly, decision makers who are equipped to turn that insight into meaningful change. That is a winning formula.