But cooking the right roles, responsibilities and skills to make up a perfect data science is not easy. Businesses might not be able to see the benefits of data science in a case the team is not structured correctly.
“An archetypal data science project team comprises Business Analysts – aka Data & Analytics Translators, Data Architects (Data Management), Data Scientists (Statistical Knowledge), ML Engineers (productionizing solutions), Platform Engineers (Foundation Setup) and Visualization Specialists (Reporting). Though diverse, data science is a human pursuit best accomplished in a team setting – the roles played can be slightly fluid around the individual strengths and weaknesses of team members,” said Prashanth Kaddi, Partner, Deloitte India.
While the basics of data science remain the same, the roles and responsibilities might differ with the industry or the scale of a business. For a company like Uber, the customer’s journey starts from opening the app, booking a trip or a delivery, and making the payment for it. But in the background there are several more processes like matching the demand and supply, pricing, tax, payments, maps analytics and many more.
“These experiments heavily rely on availability of big data around transactional and behavioural aspects of users. The high scale of our business also mandates us to automate a large part of our business processes. These automations rely on Machine learning models that are built on rich data training by using various types of user data. The key roles in our team become data engineers, business Intelligence engineers, risk analysts, data scientists and machine learning engineers.” said an Uber’s spokesperson.
Even a large spreadsheet worth of data is meaningless unless properly contextualised and presented. Teams of business analysts ensure that data from across the company is surfaced to the right teams so that they can make data-driven business decisions. Hence Business Intelligence Engineers become important for Uber.
“The second big role is risk analysts. With the complexity and scope of today’s payment systems, any business that doesn’t take adequate measures to identify and plug leaks in its systems runs a risk across multiple dimensions including compliance, financial and customer trust.. Risk analysts help us identify emerging risks and systematically mitigate them, ensuring that user experience is not compromised,” the spokesperson further mentioned.
At times, humans take longer to complete tasks when compared the capabilities of machines and algorithms to make sense of a particular dataset. Uber’s team of data scientists and machine learning engineers have fine tuned the art of developing the best models of reality that help generate decisions in such cases.
While these roles fit best for Uber’s business model. For a manufacturing company, these would be different.
Let’s assume that the scenario is to implement a machine condition monitoring system in a car manufacturing factory. There are infinite reports which can be generated from the data recorded by the sensors, but a very small subset would actually be of interest to the senior management and an extremely small subset of analysis would be so disproportionately useful to them that it will make the entire project worthwhile.
“Hence before spending any resources on a data science operation, there has to be one person who understands the business and technology and would be able to do a thorough analysis of all the processes in the business, discard everything that doesn’t need to be analysed by sanely calculating the benefit-cost ratio and then be an interface between the management and the data science team. Let’s call this person a Business Analyst or a Data Leader,” said Shishir Thakur, Co-founder, Cranberry Analytics.
The next person we need is someone who can design a data pipeline from the sensors (or any other data source) to the cloud/on-prem server (data sink). This profile is dedicatedly a Data Architect’s working in close collaboration with Data Engineers.
“In cases where the data volume and throughput are very large, we might also need a DevOps Engineer, a DataOps Engineer, and a dedicated Database Administrator. DataOps Engineers help the rest of the team with operational aspects of data flow and make the process more streamlined for continuous delivery and integration of data,” Thakur further explained.
When everything has been done with the data be it cleaning, processing and integration, at the last stage, the team needs a profile of Data Visualizer.
“Data Visualization, which is not just a part of science, it’s also an art. That’s why Data Storytelling is sometimes a separate profile. No matter how complex your analysis was, how much fidelity your sensors had and what amazing tools you invented in the process, if at the end, it’s not helping the top management with the bottom line and is not telling a clear, concise, and actionable story, it’s all in vain,” Thakur explained.
The job of a Data Visualization team is to present the final analysis in a way which is easily understandable by both technical users from other departments and the business team with zero technical know -how.
Having tales about the roles and responsibilities, it is also important to look at the soft skills of your data science team. Industry leaders believe the following are the top 3 skills to work on and master:
1, Communication: This is also a major challenge which can add a lot of value to a profile. Especially for profiles such as Data Scientist/Architect/Visualization. Everyone who needs to convey and consume cross domain information needs to be good at communication, both business communication and in person to better understand the requirements and better explain the challenges when they arise.
2, Ability to summarize their findings for leadership consumption (Business Acumen): Data scientists work with a humongous amount of data and complex algorithms which appears more like a ‘black box’ to business people and the leadership team. It’s really important that Data scientists learn how to simplify the key messaging and translate the technical findings and insights into easy to understand business messaging and actionable insights.
3, Intellectual curiosity and Dive deep: It’s very easy to be deceived by data and numbers if not interpreted and questioned in the right way. A data scientist must ask several questions on any data analysis, model performance and experiment results to ensure what data is conveying is indeed the right message and there is no error, mistake or misrepresentation. This in fact is one of the most important skills which can distinguish an average data scientist from a great one.
(With inputs from Dhrumil Dhakan)