BDMS system for Analytical Problem Solving
It was about 3weeks back when I received an email saying that I have moved a step further in my interview process with a large corporation in the US. Normally, it would have been a day to celebrate, but what I was more anxious about was that my next interview would be with a Data Scientist. A full-blown Ph.D. scholar.
For those who do not know the severity of this, they are the Navy Seals of analytics. Everybody who does anything with data wants to be called a data scientist, but only a few select make it there. So my first thought after reading that email was to introspect and ask myself if I was ready to run the data and modeling obstacle course with a data navy seal. Off-course I was ready. I had done well in my Master’s program up until now and I had solid work experience — I was set. My heart was jumping in joy to go get ice cream and then my brain kicked in and went “Doing well at school is one thing, doing well in an interview is another. You are not even 10% there, so start studying”.
So I get that ice cream tub (heart always wins), curl into my bed and open my computer to start drawing up a plan to make it through the seal training. 5 minutes later, I am fast asleep. My poor google home ends up singing “work mode” songs for me all night long.
The next morning, I wake up early and head to the library, set up my workstation for the day and do a deep dive into summarizing what I have learned, what I know and what I have internalized. I have come to recognize that the three different aspects of learnings I mentioned follow the 10% trickle-down rule. By that I mean, you know 10% of what you learn and you internalize 10% of what you know. Effectively, you internalize only 1% of what you learn. During this process of evaluating my current standing, I realised that we have learned a whole lot over the last 8 months, and it was all over the notes from different classes and projects. I needed to consolidate my learnings and hence started freestyling on my notepad and my the whiteboard and then came up with my first framework.
When I was analyzing every analytics project I had done before in my life to try and draw parallels, I recognized that these 4 parts were inherent to any data analytics problem:
- Business Context:
The process before you dive into the analysis which gives you enough information to effectively solve the problem at hand. This often includes three important parts — (a) Framing the problem statement, (b) Establishing business objectives and (c) building market/ industry driven hypotheses. - Data Extraction & Manipulation:
I call this step the garage. This is where you get your hands dirty with data for the first time, but this is also where you spend majority of the time during your analysis. This step normally includes — (a)Obtaining the data from your database, or from external sources (b) Exploratory Data Analysis where you find basic relationships within the data and (c) Data Manipulation which helps make the data worth using in the analysis. - Modelling:
Once we have the data ready to be analyzed, it is the time to decide which model to use to test your hypothesss and try and get an answer for your business problem. Up until some days back I did not really know if one could empirically decide which model to use and now I sort of do — so the process of figuring this out did quite well for me :) - Storytelling:
This is the process of converting your analysis into a story. I had read somewhere, “Storytellers are individuals who enjoy creating a holiday for the mind”. Now imagine, how impactful you can be if you can tell the end user of your analysis a story that feels like a holiday. One of the first storytelling frameworks I was introduced to was about 4 years back by my mentor. It was called the Golden Circles and it is by the the legendary Simon Sinek.
So I though I had most of the framework ready, with 6 days to go for the call. And then I showed it to one of my mentors to get his feedback and he looks at it and says “Mehhhh, there is something largely wrong with this, I will not tell you directly but I will nudge you towards it.” So he did over our conversations and that led to me morphing the 30–40….blah blah to what I now call the BDMS System for analytical problem solving. I call it a system because a system is a process and a framework is a set of tools. Solving an analytics problem is a process.
Over the next three blog posts, I aim to dive deep into each of the components of this system and how I did what I did to study for these interviews. Oh, btw — I made it past the seal training :)
About the Author: I have spent the last 5 years building a platform for student careers in India. I am currently studying the Master of Science in Business Analytics program at UC Davis to augment my business understanding with data expertise. I will graduate in August 2018 and I am currently building skills and knowledge to impact more than a billion people over the next 5 years. Feel free to write to me on sjpra@ucdavis.edu or connect with me on LinkedIn.
