Business Intelligence vs Data Science


Business Intelligence vs Data Science

Nathan Hombroek
June 9, 2017
Data Science Tags: ,
0

Welcome to the Business Analytics Arena!

Here we’ll witness the battle to determine once and for all which business analytics framework is most valuable!

In one corner we have the decades old champion, Business Intelligence, powered by software vendors like Qlik, Tableau, and Microsoft. The Champ packs a punch with interactive dashboards, curated KPIs, and detailed trend analysis.

In the opposite corner, we have the relative new comer in business analytics, Data Science. The Challenger is native to scripting languages Python and R and supported by industry powerhouses SAS and IBM. Providing powerful results with machine learning and statistical analysis, Data Science is confident in its abilities.

The contestants enter their crouched fighting stances waiting for the bell to sound….

GONG!

The arena roars to life with the audience of developers and analysts and managers and scientists shouting to encourage each champion in turn.

With the bell still echoing from the walls of the arena, the defending Champ casually strolls to the center of the ring and extends a right hand towards the Challenger. Hesitating only a moment, the Challenger exits the fighting stance and grasps hands with the Champ. Both fighters raise their hands into the air and signal an abrupt end to the bout. They leave the arena vowing to work together in harmony for the good of all. And happily ever after.

The end…

. . . Back to Reality

Nonsensical boxing matches aside, the relationship between traditional Business Intelligence and Data Science is important to note. The number of times I’ve heard someone scoff at BI dashboards or say that their business wouldn’t benefit from Data Science is difficult to count. The hesitation of some business leaders to extend their investment from one to include both is as senseless as a fictitious boxing match between business terms. After all, both analytic frameworks are striving to achieve the same result by creating a structured methodology to improve fact-based decision making with hard numbers.

Let’s walk through an example of how BI and Data Science can be leveraged together to build a powerful business application.

Accounts Receivable Case Study

Candace was recently hired as a manager to lead a team in charge of accounts receivable. She heads a team of half a dozen people who are in charge calling on overdue business-to-business invoices. Each month her team struggles to call on overdue accounts and is desperately needing a change in workflow to improve efficiency. Being the effective leader that she is, Candace immediately recognizes that she needs a way to measure the accounts that her team is responsible for so she can track her team’s progress. Candace commissions a BI dashboard to accomplish this goal.

Fig. 1: A sample accounts receivable dashboard in Qlik Sense

 

With the implementation of this dashboard, Candace is now able to track her team over time and filter receivable outcomes across multiple dimensions such as industry, region, and team member. This is Business Intelligence at its core. This view and the subsequent pages provide a holistic view of the health of the accounts receivable team and create actionable insights regarding the efficacy of each member of the team and aid in identifying patterns in non-payers. From here, Candace can identify problem accounts, manage team members, and shore up weaknesses.

While the BI dashboard has tremendously helped Candace in her day to day management and provided many key insights, it has not provided enough of a dent in her team’s massive workload. Additionally, Candace has noticed that there is a wide disparity between the efficacies of her team members. For example, John, who has been with the company for 5 years, has a much higher outcome success rate then Susan who just started last summer. Candace also recently attended an industry conference that provided her with new insights into accounts receivable. Namely, she learned that reaching out to payers prior to their invoice due dates significantly improves payment outcomes because the conversations are much less confrontational.

With these new insights, Candace is back to the drawing board with her analytics team. She mentions her learning from the conference and tells the story of John’s success compared to the rest of the team indicating that his experience gives him the ability to identify the accounts most likely to achieve a positive outcome. This ammunition in hand, the analytics team develops a plan to create a machine learning algorithm that predicts the payment outcome of an invoice prior to its due date. They outline the problem statement and collect data regarding both the invoice details and the account history and demographics. Their goal is to provide the tools that enable every team member to be at least as effective as John.

Fig. 2: A Decision Tree written with R implemented in Qlik Sense

With an algorithm developed, the analytics team creates a basic model that they can take back to Candace and her team to display how the Data Science model framework operates. By creating a visual that the team can follow and understand, the analytics team is able to build the trust necessary to convince the accounts receivable team that a change in workflow will be beneficial.

In this case, the analytics team displayed a pruned Decision Tree classification model with the explanation that a collection of similar trees are used in the Random Forest classification model that is actually predicting the outcome of invoices. The algorithm classifies every outstanding invoice and predicts whether a payment will be On Time, Late, or Unpaid as soon as a new invoice is written. With the trust built and the model delivered, Candace now has the ability to introduce a new advanced workflow that her team can follow to achieve drastically improved payment results. The key to the success of the new Data Science enabled workflow is that it allows for a seamless transition by providing immediately actionable intelligence to Candace’s team by generating a call list with phone numbers and links back to the CRM source system. Here, the call list is ordered by the payment prediction and the outstanding invoice amount.

Fig. 3: A Data Science enhanced AR Workflow implemented in Qlik Sense. The columns that allows the team to prioritize their workflow are highlighted. (#’s and links are anonymized)

Candace has now achieved the endgame of her business analytics journey and has dramatically improved her team’s ability to maximize the cash flow of her business.

Conclusion

With this case study, we saw how a common business problem can be approached with both Business Intelligence and Data Science together. This team now has the ability to both accurately track the history of their efforts and predict future outcomes. The marriage of these two frameworks leads to an optimal fact-based decision making process that can be applied to nearly any business problem.

What ways could you leverage BI with Data Science in your business to improve outcomes?

Are you doing it already?

Share in the comments!

 

Nathan is an experienced analytics professional with in-depth knowledge of data science, business analytics, and business intelligence. He leads the Data Science Practice at Axis Group that focuses on solving business problems by providing business partners powerful, data-driven solutions and actionable insights by leveraging advanced statistical models, machine learning, operations research, visualization, and business expertise. Nathan is a two-time graduate of Georgia Tech with a BS in Economics and an MBA with a concentration in Business Analytics.

Leave a Reply

Your email address will not be published. Required fields are marked *