With each new instance, I'm reminded of a detective sifting through data in search of a target for which there are no leads or traces. Every new day at work feels like a fresh episode of some detective web series, starting with an attempt to comprehend the data, its dimensionality, key attributes, and their relevance in predicting the outcome. Every day reveals a new piece of information about the data, creating excitement for the next day to arrive to the office with a different perspective.
Conversely, talking with business executives to gain a better grasp of the data or surge of new data might appear pointless at times owing to their own thinking processes and lack of recognizing the importance of data. Their inability to comprehend and urge to impose norms results in a never-ending conversation. It is usually tough to persuade businesses to use data that they possess. They won't be convinced unless they see a business benefit from the expected outcomes - which is very true! Regardless of how fantastic the anticipated results are, business value is the ultimate goal for them. As a detective, I'm always afraid and aiming for outcomes that provide real value to the company.
Here are some tips to be a successful data science detective:
- Understand the case before you start working on it, understand the target business value (Business Understanding)
- Spend more time to find right clues (Data Understanding)
- Prepare the clues (data elements), assign weights and importance (Data Preparation)
- Choose multiple detective methods (Modeling)
- Exercise all but apply the one which provides best business value (Evaluation)