Data Visualization: A Bridge Connecting Data Scientists and Business Leaders

By Cesar Koirala, Director, Data Science (Talent Analytics), Liberty Mutual Insurance

Cesar Koirala, Director, Data Science (Talent Analytics), Liberty Mutual Insurance

Today, data science is being rapidly adopted by the business world; sales is utilizing advanced algorithms to predict revenues; Human Resource relies on natural language processing to hire better candidates; IT is improving your web experience using machine learning. While this rapid adoption of data science is encouraging, I believe it also needs caution on part of both business leaders and data scientists, because we do not speak each other’s language, yet!

A new data scientist is proficient in math, coding, and algorithms. A business leader on the other hand is concerned about strategies, planning, and outcomes, usually tied to measurable units like cost and revenue. This difference often makes is difficult for the business leaders to properly communicate their business goals to data scientists. Alternatively, data scientists often struggle to demonstrate how their approaches and results meet those business goals. This is where data visualization can be very resourceful. If used properly, data visualization can translate between the vocabularies of business leaders and data scientists thereby removing ambiguities.

Let’s take an example of Human Resources (HR) to illustrate this point. While HR vocabulary has jargons like compensation, benefits, and attrition, HR data scientists think in terms of algorithms and probability distributions. By using data visualization, HR teams could close the communication gap in two important areas.

1. Proper translation of business goal into data science objective
2. Transformation of complex data science results into actionable business insights

 If used properly, data visualization can translate between the vocabularies of business leaders and data scientists thereby removing ambiguities 

A data Science project begins with a business goal, and the first step in the project is to translate that business goal into data science objective – what metrics to consider and what algorithms to use? Data scientists often engage in an activity commonly known as exploratory analysis during this process. This could be something as simple as summarizing the data numerically. Although simple, such exploration can be extremely powerful. The charts and graphs based on data exploration allow the teams to see trends and patterns in the data. They are not only able to visualize the factors that affect the business goal, but also able to understand relationships between these factors and their associations with the business goals. I revisit the HR example below to illustrate how HR teams can utilize data visualization to identify factors impacting their business goal of reducing employee attrition.

In this hypothetical example, HR is able to identify that performance rating and commute distance are correlated to employee attrition. More specifically, attrition decreases with rise in performance rating and it increases when commute distance rises. 

Complete understanding of business goal and the factors impacting it enables data scientists to choose the best data science tool for the task in hand. This, in turn, enables the teams to close the communication gaps when translating business goals to data science objectives. In our hypothetical example, HR data scientists should discuss the results of data exploration with business leaders (domain experts) to make sure their understanding is aligned.

After complete understanding of business goal, data scientists often engage in an activity called modeling. Models are mathematical representations of real world situations. Models could simply be explanatory in nature helping to understand what caused the situation. Going back to HR attrition example, if HR data scientists construct a regression model, they are able to better understand the underlying factors that contribute to employee attrition. Models could also be predictive in nature. Such models learn from present and past data to predict future outcomes. For example, attrition predictive model could predict future attrition rates based on historical attrition data.

Although models are complex in nature, their results do not need to be complex. Data visualization can be used to transform model results into useful business insights. This, in turn, enables leaders to make informed business decisions.

Dashboards are often used to make models more accessible. They are interactive business intelligence tool that can be used to visualize model metrics as well as model results in intuitive fashion. Visualization of model metrics can help business leaders understand what factors affect the business goals and to what extent. Visualization of model predictions can aid in critical business decisions.

Data visualization was already an important part of the business world. Integration of data science in to this world has opened up an additional opportunity to utilize it effectively. By visually communicating metrics related to business goals and results of modeling, data visualization can be a bridge between business leaders and data scientists.

Read Also

Clinical Trial Technology: Drive Adoption with People

Clinical Trial Technology: Drive Adoption with People

Dawn Niccum, Sr Director, Quality Assurance & Compliance, inSeptionGroup
Market Wisdom Comes With Age

Market Wisdom Comes With Age

Margo Weisz, Partner, City Lights Group
Sharing Expertise through Telemedicine

Sharing Expertise through Telemedicine

Todd J. Vento, Medical Director for the Infectious Diseases Telehealth Program, Intermountain Healthcare

Weekly Brief