What made you venture into analytics in the first place?
One of the focuses in my graduation was operations research. Optimization, simulation and game theory were part of the course. While learning, I never expected that someday I would be using some part of it where I work. Most of the people were joining as software engineers. At that time when analytics was not a prominent branch, a small start-up in analytics came to the campus and they had a very different unique interview which tested the knowledge of the person in game theory, probability and mathematics which I liked and I joined the firm. This is how I began my career in analytics.
What are the challenges you faced and how did you overcome? How much did you enjoy the work? How was the overall journey?
At the beginning of my career, I started with analytical tools like SAP and SPSS. I started enjoying it because I loved those equations and probability. At the end of the day, when I was running logistic regression or linear regression and when I was predicting something and when I was forecasting something, I could relate to those things.
When I went for my masters in Arizona, I discovered a whole new side of it. There were many new tools which I was not aware of - like Python, R-programming, which are excellent in machine learning, unsupervised algorithms, and support vectors and so on. These gave whole new dimension to the normal data mining which I did extract that time. Since I was independent consultant for various pharmaceutical companies and also for US Government, I kept discovering new things and new applications of either same models or different models. My range of models kept increasing from neural networks to support engines and in past 2 – 3 years, I was able to learn and implement many things in social media analytics. As of now, I have a team and I interact with statisticians, mathematicians as well as programmers on a daily basis who come up with new concepts every day. I verify those concepts, check their feasibility and try to implement them. Now it is more of advising people on analytics as opposed to previous time where very few people were there to advice. It was more of either pure statisticians or pure business, but no bridge in between.
What are the main qualities a person should have in him to get into analytics?
There are two ways a person to get into analytics – By having an Inclination towards statistics or numbers, learning models would become almost easy with this inclination; other one is as a business analyst, having a strong inclination towards understanding business processes. Even though the statistical models are really good, they are not always good at showing the actual ROI. There is no meaning to the work at the end of the day if you do not show the actual ROI. The higher management needs to see the benefit in that. Someone with business standards can help you with that. So, for any project on analytics it is necessary to have these two kinds of people.
What are the aspects which he can learn academically through different courses? How is the learning curve over the years?
In analytics, people start with tools like logistic regression, linear regression, SAP, SPSS for one year and move to decision trees, neural networks, clustering etc. It takes a lot of practice and 2-3 years to master them. Just taking a course won’t help you learn anything unless you get your hands dirty with data, face a lot of problems. When you want to switch domains, your speed of learning becomes faster since many of the models are similar across the domains. For example, I might use survival analysis for employee attrition rate as well as patient risk rate. The model remains same, but the business application may differ in HR and health care. So, in the beginning 2 – 3 years the learning curve is flatter where you need a continuous concentrated learning and from there the learning curve is much sharper.
Different models you are talking right now are taught as part of curriculum, how can we use them in pursuing the career in analytics? The awareness of the career in analytics is very less. What do you think is the scope and career opportunities in analytics today?
When you learning Java, C programming and other stuff in your college, you are taught on how these languages behave. Only when you enter a company, you are taught on how to apply them and after 3 - 4 months of practising, you become a master in that and you work on it for years. Analytics is pretty much the same, where you learn the tools in college and apply them during the work.
As for the second part is concerned, Analytics has emerged as an industry of its own. There are many parts coming up. There are people who act as a bridge between statisticians and business people; they are called business consultants or business analysts. Some concentrate on research side; they are called data scientists. There are people who concentrate on programme management or project management; they are called programme managers. There are people who have their own innovation labs for analytics in their organizations to suffice the need. Analytics are of two kinds. One which is easily digested by the businesses – propensity to pay models which includes logistics regression, linear regression or time series models. Some things like neural networks, support vectors act like a black box. It is very difficult to explain to business people what happens inside the model. For the second type of analytics, people have their own R&D sessions and innovation labs. As analytics keep evolving new branches might come up and old ones might merge into one. Being in analytics is itself a major learning opportunity and unless you enter it you won’t know what is in store for you.
What prompted you to write your books? What was the inspiration?
I’ve written three books so far - Social media analytics targeted for analysts; social media visualization targeted for designers and visualizers; and social network topology and Ricci flow written specifically for mathematicians. The reason I wrote these books because I felt that the way we look at social media and social network and analyse them, doesn’t do justice for the kind of benefits we can get from them. Just by doing the sentiment analysis or text analysis or comment analysis and showing to people does not empower them. It is all worth when you predict whether a video will go viral or not. Many may say content may be the main reason for virality, but it is not as simple as that. There might be lot of catalysts and promoting factors in your network which you need to identify. You choose certain path and connect some points in order to make your video viral. Even you use the advertising in Facebook, Google and pop-ups which you use to make your conversion rates high. But, it is not that high because somewhere down the line we are stuck up with the analytical models we started using for other domains. They might not work for social media. We just don’t need small tweaks; we may need entirely new models. It’s far away from regular regression, may be one neural network for that matter. I did some research and wrote the books for this reason. Also it is difficult to switch your domain in a firm which is not a complete analytics related. So, I wanted to write these books so that I can have a comprehensive exhaustive material online which anybody can use and which covers all the things about social media analytics which has been done and which can be done.
Are you planning to write another book? What are your future plans?
I have been having email exchanges/discussions with mathematicians who are attending field prize ceremony and giving lecture at the same ceremony. I’m thinking to write a part II for the social network topology.
Note : To contact Prasad Kothari, you can reach us at email@example.com