If you’ve spent some time in the tech industry, you may have come across some folks with this intriguing job role: Data Scientist. Touted as one of the hottest jobs in the tech space, a data scientist is one tasked with making sense of the reams of data an organisation filters in. These random bits of data need to be strung together in meaningful patterns, leading them to generate some precise insights on various metrics like consumer behaviour, expenditure patterns etc. This kind of personalised data funnels can significantly help multiple organisational functions shore up their strategies like marketing, sales, business development and more.

With organisations going the data-first way today, a data scientist happens to be one of the most important people on their payroll. Let’s find out what it really takes to be one today:

Cindy Mathew is a consultant at TheMathCompany, a hybrid consulting firm that builds contextual AI assets for Fortune 500 companies. At TheMathCompany, she works with the data sciences team of a leading US retailer to build and deploy scalable applications– leading various projects ranging from data pipelining, data curation, performance measurement and prediction of marketing initiatives. “My main responsibility is translating a customer’s business requirements into actionable steps that can be coded out, and brought together as a complete solution by each team. It is pretty exciting to see how our every-day work – the codes we write, the reports we create etc. – translate into actual business impact.”

(Cindy Mathew, consultant, TheMathCompany)

With a passion for numbers and patterns, Mathew says everything can be broken down into numbers and equations, if you look at it the right way. “The 2008 recession was predicted by observing patterns in the data. The ongoing pandemic was predicted last year based on patterns extrapolated from data. Current methodologies structured on obtaining sustainable energy can be improved through the usage of new algorithms, image recognition and more.”

AM To PM:

A typical day for Mathew begins with “scrums” with individual teams – this is a review of the previous day’s work, planning out action items and discussing potential roadblocks in delivery. There is a keen eye on mistakes being made too, giving other teams a chance to watch and learn, and allowing for remediation plans. A big chunk of the day goes into helping the team out with solution design, decoding problems for analysts to work on and quality checks. “You need to be able to switch tracks across different projects at a rapid pace through the day, and be ready to get your hands dirty as and when required,” says Mathew. Working with a company like TheMathCompany, which caters to a range of clients across domains, the challenge is to obtain technical and domain-specific context of different problems. Thanks to a diverse ecosystem with professionals of varied experience, peer reviews and cross learning is a daily affair, enriching the quality of ‘solutioning’ across the board. “It is imperative to understand what questions the customer wants to answer using his data. This helps us contextualize solutions, while re-using the internal assets we have in our organization to speed up the process.”

Prior to a project commencing, Mathew has a pretty meticulous checklist she endeavours to follow. The primary mantra? Understand the end goal of a project and work backwards.

“Many analytical efforts fail because the end results are eventually not consumed – this can be because the results do not solve the right problems or are in a format that is difficult to be consumed. Building a solution while keeping the end consumption in mind is extremely important,” she adds.

Other important questions she ponders over are the business outcomes of the project that usually entail increasing RoI, pre-empting behaviour or improve existing processes; nature and manner of actions being taken to reach this outcome; creating assets to achieve these goals like dashboards, models, decks with data and more; measuring success of the project and finalising appropriate metrics. “In the initial stages, it helps to have a high-level view of the complete project flow,” she explains.

An additional hygiene factor is to consult with colleagues who may have worked on similar projects, and do a quick assessment of resources available on hand to execute the project.

Not Just Coding; But Problem Solving Skills Lie At The Core of Sound Data Science

From figuring the right way to handle data to identifying the most appropriate algorithms to use and iterating on optimal solutions for the best results, a data scientist has to be prepared. Her past experience as a data scientist in a similar services environment helps Mathew manage challenges, and is grateful to her evolution from “number crunching” and coding; to looking at the bigger picture, and approaching data from the lens of problem solving. “While the nature of challenges remains the same, the approach to problem solving in services has evolved over time. Looking at a business problem holistically rather than only from the perspective of data, spending sufficient time on solution design rather than jumping straight into the numbers and coding, and thinking outside the box for solutions, instead of sticking to time-tested approaches only – this evolution is the result of an ecosystem where the focus is on creative problem-solving.”

Given that this is a relatively new field, and efforts to upgrade college curriculum are ongoing, Mathew recommends professionals to pick up skills along the way. In some cases, L&D divisions within companies strive to educate employees on industry-relevant skills. “At TheMathCompany, there is a lot of focus on continuous learning. A dedicated in-house growth accelerator team constantly pushes out organisation-wide learning programs in different formats as well as learning materials that span domains and industries. We also have InfoShots – a series of sessions where anyone in the org can conduct short learning sessions about pretty much anything (new algorithms, business domains, how to use new tools/platforms etc.)

Ask Questions, Probe the Data and Never Lose Sight of the Big Picture

Translating the rows and columns of a table into what they actually mean and represent is the biggest challenge. We often tend to get lost in just numbers and aggregations, and sometimes lose sight of what they mean for the business. It is important to visualize datasets to understand what each row represents, what each column means and the information they add on to the numbers. Measuring data quality accurately is the other challenge we often encounter – how representative are the numbers in front of you of the real-world? Business understanding has to be layered on to the numbers we see and datasets curated in the right manner, before plugging them into a standard algorithm/model. Finally, asking the right questions is key – dig deep into the dimensions and metrics in front of you, tie it back to the real-world and question whether it all adds up

With automation becoming a core tenet of business today and organisations catapulting into digital transformation, the emphasis on problem solving is critical to power a company ahead. “Along with developing AI, humanizing and translating AI solutions to address every-day problems will be a big part of the future,” she signs off.