<My first post on LinkedIn after the invitation to publish>
The other day, I was watching a fascinating soccer match (English Premier League). Somewhere towards the end of the game, I realized that the poor fellas had been running for about 94 minutes to end with a 1-0 scoreline. Somewhere, in those 94 minutes, there were more than 12 genuine attempts at scoring, several high quality defense maneuvers, many strategic plays, but just one that hit the bulls-eye. Was it worth it? Yes. Were the remaining 93 minutes 30 seconds irrelevant? No.
To me, those 93.5 minutes, and the quality of play therein, decides the quality of the team, the real value of that 1-0 scoreline. That grunt work decides whether it was all worth it in the end. Agreed, in the end, a 0-0 scoreline is not even half as awesome as a 1-0 scoreline (especially when my (preferred) team wins). But would I rather just watch the highlights (and that one goal) along with the match ending analysis? Hell, no!
Fortunately for my clients, and unfortunately for me (and maybe some of my team members), at a very early stage of my analytics career, I graduated from being an individual contributor to being a team and engagement manager. The clients never had to realize how sub-par I was at being an individual contributor (one deficiency less!) while my teams had to dumb things down for me at times. However, a common emerging theme through the years and projects has been the power of grunt work – the act of doing the supposedly less interesting tedious work over and over again. We tend to discount the tedious work for the glory of the big moments.Having to clean and reorganize the data a few times over in a project to ensure that the business objectives are still central, and not the data scientists’ need to build a super-complex model or deploy a super-efficient algorithm. The highlights look better focusing on the buzzer beaters than on the continuous court runs required to be at the right place at the right time for that buzzer beater.
What do I mean by that?
In any form of sports, coaches insist on getting the basics right. The stance, the form, the shot, the power, the follow through, the stable head position, and so on. And to get these right, a good player (and not even a great player) has to keep doing it again and again. Borrowing an analogy from the game of Cricket, I have always been fascinated by how after getting beaten by a particularly beautiful delivery, even great batsmen like Sachin Tendulkar or Rahul Dravid or Ricky Ponting play out the same delivery in their head and in their shadow practice a few times. One can see them going through the motions of replaying the shot they had just played and a calculation of sorts on what they would do differently the next time.
Having reached the level of proficiency they had already, they shouldn’t be bothered with it, right? Wrong! You still have to go out there and get the basics right. Every time. All the time.
Likewise, in any analytics project that we undertake, one of the most critical things is to get the baseline right. The problem that we are solving, the exact analytical problem definition, the data that we’re going to look at, the process of reformatting and re-purposing the data to our needs, the need to do the univariates and bivariates before we get to the multivariates and the models, and so on.
And it is here that one of the most glaring needs of the analytics and data sciences industry lies, in terms of talent creation and talent management. The way we are growing our data scientists of the future, we are hoping that most of them would be able to build a model without touching and feeling the data, or find AI/NN solutions without caring about the basic business hypotheses.
And be honest here! How many of us find this dealing with the data business one of the most tedious things to do? Something that we would happily outsource to someone else? Something that we would happily let someone else take the accountability for? Or blame for?
What Ticks In Analytics 