UPDATED: Feb-2. As organisations increasingly harness insights from zettabyte haystacks of data, the demand for data scientists and analytics specialists continues to soar. Companies have invested billions of dollars in data science talent and fancy analytics tools. The investment has paid off, to some extent. But data scientists, who are perhaps overrated, have been falling behind, in terms of business skills. A Kaggle’s 2017 survey of data scientists reveals that the issues are not inadequate technical skills. It seems to be a last mile problem – how data scientists communicate the insights and the analysis back to the non-technical business person. It’s also about the ability to understand the problem statements of the business.
To ascertain this and get more context, DIGITAL CREED spoke to Rangarajan Vasudevan, CEO, The Data Team. Rangarajan or ‘Ranga’ as he is fondly called, runs a successful data science company with offices in Chennai, Bengaluru and Singapore.
Excerpts from the interview:
DC: What are the critical skills that a Data Scientist needs to have today?
Ranga: The critical skill set that goes beyond technology and algorithms and stuff that is talked about in coursework, is the ability to understand the business. It is about the ability to understand what problem statements make sense in that business, and how the solutions that are created from the data science process can be applied back to the business. That ability to translate is a skill lacking in many data science graduates. That’s why businesses have not been able to consume the insights in a meaningful way.
There is a related point. Ultimately, whatever data science creates does not exist in a vacuum. You need to embed that operationally as part of a business process.
Otherwise, it becomes just a boardroom discussion or PowerPoint presentation without any physical impact. The operational process change is change management that the business must undertake. That is something outside the ambit of the data science world. The business itself needs to be prepared for change management. Both go hand in hand.
DC: Do you agree that all this investment in data scientists is not yielding results? What needs to be done to correct this?
Ranga: I agree with you. The investment is not paying off. The way that it is being done today — many companies make the mistake of just hiring people with a math background or a computer science background, and feel it is enough.
Until the supply of this skill becomes large enough, there will always be well-funded startups who will come in. The talent will go to the startups who are well funded.
The impact is also well understood. The data scientists build a model and it gets immediately applied to the startups. They will know the impact within the next week itself.
In a large organisation, the change management itself takes months.
DC: What are the essential concepts that a course on data science should include as part of the curriculum?
Ranga: What works well is the combination of the right teaching methodology and hands-on experience. In many cases, some of the coursework puts emphasis on pedagogy. Learning things academically, without hands-on application. That becomes counterproductive, especially for data science.
It is not just the content or the libraries of software, that are important, but the ability to see that in action is a very critical piece. We need more collaboration with industry, academia, and startups. It will provide very good problem statements, and offer good recruitment opportunities. Data sets and models can be shared. This will make the coursework more tangible.
On the other side, it is rare to see a weekend course set up working very well. If a person were to dedicate time to learning — say a six-week residential crash course — or a 1 year MBA — those type of programs have a much better return on investment for a data scientist — as opposed to doing it once every weekend. That will not make you employable.
Choose those programs that include hands-on applications and which are aligned to industry partners.
DC: What are the technical concepts to look out for?
Ranga: The important thing to always consider is not the algorithms for the modelling but the modelling process. The process of data science is a lot more important than the algorithms. If the course can teach the process as opposed to teaching individual algorithms that will always be better. Because you know how to apply something, as opposed to just learning the algorithms that are going to be applied for this kind of problem statement.
In the real world, there will be a lot of changes in the way data is captured, that will have an impact, and this algorithm may not be suitable.
So the data science process is very important.
DC: Where do you see demand for Data Scientists in India? Which industries are hungry for data scientists now?
Ranga: Every industry is waking up to the possibility. Earlier you could restrict it to the well-funded startups and the BFSI sector. But 2018 has been a watershed year for manufacturing waking up to the possibilities of AI.
When traditional industries like manufacturing start to adopt this… Most of the large industrial groups in India — when they wake up to the possibility — then it will have a ripple effect on all their holding companies, which means pretty much every sector — real estate, construction, cities, highways, — everything has a direct bearing. Of course, the government is also do its bit by talking about Digital India, Startup India and other initiatives.
I don’t think it is restricted to any industry like it used to be 2 – 3 years ago.