One evening during a trip to Ranchi, a small town in India, where I was born and grew up, I was sitting on the side of a small lake, away from the cacophony and glitter of the New York avenues and rooftop lounges, that I was struck by an epiphany! I realized that the fundamentals and values instilled in me during my growing up years were the reason I could adapt and excel in the modern cosmopolitan world driven by trends and fads.
On similar lines, in the analytics world, if we peel away the buzz words like ‘big data’, ‘machine learning’, and ‘internet of things’, what remains behind every successful data modernization program is a good old ‘data model’. If you focus on the design of the core data model to make it adaptable to the ever-changing business models and needs of an organization, you will be left with successful transformations programs.
I forayed into analytics when the CEO of an e-commerce company in New York City who I was consulting for pulled me into his office and said, “I want you to build a data strategy for our company”. Despite several years of handling business process re-engineering and other digital transformation programs for large companies, I knew that building a data strategy was a whole different game. I had to start from scratch, analyze the firm from a completely new lens, and also unlearn some of my learnings from my previous experience, and venture into the unknown. The gamble paid off and it eventually turned out to be one of the most successful transformation programs within the organization encompassing data governance & management, platform modernization, and data monetization topics. Since then, over the years, I have been a part of several enterprise data analytics transformation programs which have been some of the most rewarding experiences for me both professionally and personally.
Sometime back, when I was reading extensively about the psychology of data analysis; I stumbled upon the ‘Concept of Biases in Data Analysis’ such as Confirmation Bias and Selection Bias. In essence, it refers to a phenomenon of subconsciously looking for insights that either confirm with their assumptions or uses selective datasets to prove their hypothesis. On looking back at some of my work, I found them contaminated by one or more of such biases. I was shocked to know I had been essentially using data to prove what I wanted to see rather than looking for what the data really wanted to tell me. The problem didn’t end there and I found myself exhibiting similar biases in the real world as well when interacting with friends and family. I used to listen and respond to only what I wanted to hear rather than what people were actually trying to convey. I then started taking conscious effort to get rid myself of these biases both from my work as well as my social conversations.
To be a true data scientist, it is imperative to listen to what the data wants to tell you rather than finding answers to only those questions that need answering. A large portion of the work in data analytics and reporting is essentially storytelling and how well you can tie various aspects of data to convey using simple and concise visualizations. I don’t believe in ‘loud’ reports that have too much going around in them as they are often distracting and don’t add any value.
Meet Apurva Pandey: Mountain Biker, Tennis Freak, and AI Enthusiast