Bala Deshpande is a thought leader in the manufacturing analytics and data science space. He founded and sold a data science company. Also, wrote the book “Predictive Analytics and Data Mining: Concepts and Practice with RapidMiner.” I don’t know how he found some time, but he also started the Predictive Analytics World — Manufacturing conference.
I connected with him in Ann Arbor, Michigan, while searching for mentors in the data science space. He always was gracious with his time and gave me a lot of good advice. In this post, he kindly responded to a few questions I asked him about analytics.
How did you get started in analytics?
Coming from an engineering and experimental background, I had the opportunity to get involved with data and analysis during my graduate school days. Back then (early 90s – long before the present day hype) “analytics” meant collecting raw data, cleaning and transforming it using clunky workstation based spreadsheets (heard of Lotus 1-2-3?), fitting regression models and interpreting the model coefficients to see if it made engineering sense.
… the basic spirit of analytics has not changed: convert data into meaningful and actionable decisions.
What excites you the most in analytics?
While the tools and technologies have advanced to incredible levels (back in the day, fitting a model on 1000 data points with 20 or 30 attributes meant an entire weekend of computing!) the basic spirit of analytics has not changed: convert data into meaningful and actionable decisions. The thrill of taking a bunch of raw (seemingly random) numbers and converting it into a beautiful and intuitive chart or building a model to then predict what is likely to happen, still remains the same.
Can you give some exciting or unusual examples of application of predictive analytics?
Taking the images from fabric designs (patterns) and extracting features such as major color, geometries, textures etc in order to determine if there is an underlying pattern (no pun intended!) to the sales success or failure of products, based on region, based on season is one really exciting area that we have worked on recently.
What are some current trends in analytics?
We see a lot of niche verticals springing up which have suddenly “discovered” analytics. Legal (should we fight or settle this case?), Restaurants (buying commodities on contract or spot prices), Medical (cosmetic surgery – a good candidate for nose job or not?) and of course sensor data related analytics. The whole space of IoT is opening up.
Where is predictive analytics heading?
Analytics for the last 20 years or so was purely on structured data: tables of rows and columns of numbers. Recently text analytics has started to rapidly expand and the future seems to be headed to more unstructured data analytics: more text, images, audio/speech, video, etc. However in my opinion the volume of actionable analytics will continue to be driven by structured data coming from sensors and machines.
What challenges do you see for analytics practitioners in the future?
Keeping up with new tools and technologies. However in 3-5 years today’s hottest technologies will be heavily commoditized, automated or obsolete. How do you balance your precious time between learning the latest trend and delivering today’s goals is a big challenge.
How do you balance your precious time between learning the latest trend and delivering today’s goals is a big challenge
With the arrival of analytics as a service, do you see the demand of analytics professionals go up or down? Why?
I don’t think self serve analytics will reduce the demand for analytics professionals. Ten years ago one may had to write several lines of lines code to get data from one shape to another, today there are tools to automate that. But the person running the tool (or writing the code) are still the same. Much of the drudgery will go away, but the business of asking questions, determining which data can help to answer these questions and understanding and applying the results of an analysis will still require a professional.
There is one caveat though: as more and more niche applications come up, these tend to get heavily automated and machine learning may take away many of tasks which a professional is today spending time on.
What one advice will you give to aspiring analytics professionals?
Always ask questions at every step of the analytics process. Try to keep assumptions to a minimum, but don’t ever ignore or lose sight of these assumptions.
Question everythingAlways ask questions at every step of the analytics process. Try to keep assumptions to a minimum, but don’t ever ignore or lose sight of these assumptions.
Download this Free Guide to become a data scientist
Get this FREE self-study guide as a beautiful, high quality PDF.