Data Science in a Business Context: Achieving the Perfect Balance

Blog Post created by michelleadvaney@fico.com on Jun 5, 2017

Data science is often misunderstood. We’re constantly hearing terms like Big Data, analytics, artificial intelligence, and machine learning, but how can these trendy technologies solve actual business problems?


Before I jump into my blog about the power that comes with integrating the results of data science into an enterprise strategy, let’s observe a modest conversation between a data scientist and a business owner. Now imagine, you’re at the water cooler, and you hear:


Data Scientist: Now that we have the Hadoop cluster set up, we can finally amalgamate our historical data.

Business Owner: Terrific! But, uh, what are we supposed to do with that?

Data Scientist: It’s really cool, with this large of a dataset we can apply machine learning, and build out regression models, and…

Business Owner: Ok, but what I really need is to understand what campaign to run to lower shopping cart abandonment and increase our revenue. How are you going to help me do that?


As businesses adapt to become ‘data-driven,’ as so many are, there is often a disconnect about how that data can actually be used. Data scientists are typically siloed in their world of algorithms and wrangling while the people driving business decisions are searching for answers to increase revenue and profits. While the data scientists are sitting on a potential gold mine of business insights, it can be hard to map the results of data science to improving the bottom line.


This challenge persists in organizations across industries, but admitting the problem is the first step. Some schools are starting to recognize the need for companies to marry business with data science; Duke offers a Masters of Quantitative Management Program at the Fuqua School of Business. This program is built to specifically address the importance of using data to derive business value. The Assistant Dean of this program, Jeremy Petranka, believes that “being able to organize and manipulate data is rapidly becoming commoditized” in the marketplace. Data is valuable not for data’s sake, but for the ability to discover “actionable insights” to address “key strategic questions.”


ai aml.jpg

How do we bridge this gap to find these coveted ‘actionable insights’? Adhere to the following principles:


Know what you want, and what you need to get it.

Understand the business imperative, and then understand the business impact of your decisions, and what you hope to gain from them. Once you can clearly articulate the questions that need answers, people with whom you work will be able to better apply analytics to solve for those specific business challenges.


Based on the industry standard Decision Model and Notation (DMN), tools like FICO® DMN Modeler (part of FICO Decision Management Suite) outline a decision flow overview. This tool helps to drive a basic understanding of the knowledge sources, inputs, and logic required to make a decision. Users can then identify what supporting data is necessary to make a final decision. DMN Modeler takes a decision-first approach, which helps users understand the components in the decision they’re trying to make. This allows for transparency in the decision process, simplifying the collaboration and efficiency by clearly mapping everything that goes into making the decision.


Identify and gather the data.

Where does this data come from? What form is it in? Is it from social media? Is it from consumer purchases? Is it someone’s credit history or background information? Where the data and information comes from can help you understand what type of analytic insights can be gained from it. Processing that data is the next step in taking the information and making it meaningful.


When considering knowledge sources and inputs, you should assess all available data and whether or not it has already been analyzed. Scorecards, for example, can be used as inputs into decisioning to achieve a sophisticated level of analytic insights. Companies can also use data from external outlets to complement their own analysis and services. For example, external scorecards can be used in decision models to better predict behaviors, providing a more accurate decision outcome. However, it is important to keep in mind that data or analytics may need to be further processed (especially if it’s in an unstructured format), where capabilities like text analysis and natural language processing come into play.


Visualize the ideal outcome.

Many tools provide the ability to visualize analytics through graphs and charts. In fact, there are countless ways to uniquely represent the same information visually. For this reason, it is important to present the information in a way that clearly communicates what data is available, and what results are expected from that data. Presenting information in such a way allows data scientists to understand what is expected of the data and allows them to manipulate information according to business goals.


Tableau for FICO® is integrated in the FICO Decision Management Suite, making it easy for business users to visualize data, decisions, and the effectiveness of applications. This eliminates friction between organizations (like data scientists and executives) by unifying goals and outputs.


Terms like Big Data and machine learning may be all the rage right now, but the real value lies in understanding how these technologies can be applied in conjunction to achieve better business outcomes. When done right, bridging the gap between data scientists and business owners results in better strategic planning and better results for the company and individual teams.


This gap is already being bridged! To learn more, check out the links below:

Duke’s Quantitative Management Program

FICO’s Academic Engagement Program

FICO Decision Management Suite